BasicModels.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "WARNING: Logging before flag parsing goes to stderr.\n",
      "E1012 23:24:23.536626  6832 ioloop.py:763] Exception in callback functools.partial(<function Kernel.enter_eventloop.<locals>.advance_eventloop at 0x000001AA5B7FF048>)\n",
      "Traceback (most recent call last):\n",
      "  File \"c:\\users\\jacob\\appdata\\local\\programs\\python\\python37\\lib\\site-packages\\tornado\\ioloop.py\", line 743, in _run_callback\n",
      "    ret = callback()\n",
      "  File \"c:\\users\\jacob\\appdata\\local\\programs\\python\\python37\\lib\\site-packages\\ipykernel\\kernelbase.py\", line 306, in advance_eventloop\n",
      "    eventloop(self)\n",
      "  File \"c:\\users\\jacob\\appdata\\local\\programs\\python\\python37\\lib\\site-packages\\ipykernel\\eventloops.py\", line 238, in loop_tk\n",
      "    app.tk.createfilehandler(stream.getsockopt(zmq.FD), READABLE, notifier)\n",
      "AttributeError: '_tkinter.tkapp' object has no attribute 'createfilehandler'\n"
     ]
    }
   ],
   "source": [
    "import sys\n",
    "import time\n",
    "import os\n",
    "import numpy as np\n",
    "import tensorflow as tf\n",
    "from tensorflow import keras\n",
    "from sklearn.model_selection import train_test_split\n",
    "from sklearn.neighbors import KNeighborsClassifier\n",
    "from sklearn.naive_bayes import GaussianNB\n",
    "import matplotlib\n",
    "matplotlib.use('TKAgg')\n",
    "from matplotlib import pyplot\n",
    "import matplotlib.pyplot as plt\n",
    "from sklearn.metrics import auc, roc_curve\n",
    "from sklearn import metrics, svm\n",
    "from sklearn.metrics import precision_recall_fscore_support\n",
    "import pandas as pd\n",
    "from sklearn import preprocessing\n",
    "import warnings\n",
    "warnings.filterwarnings('ignore')\n",
    "from sklearn.utils import shuffle\n",
    "from tensorflow.keras.callbacks import TensorBoard\n",
    "\n",
    "def dataProc(X1,y1):\n",
    "    X1 = pd.DataFrame(X1)\n",
    "    X1=X1.to_numpy()\n",
    "    y1 = y1.to_numpy()\n",
    "    \n",
    "    \n",
    "    min_max_scaler = preprocessing.MinMaxScaler()\n",
    "    X1 = min_max_scaler.fit_transform(X1)\n",
    "    \n",
    "    X1, y1 = shuffle(X1, y1, random_state=0)\n",
    "    return X1,y1"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Loading and Processing Data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "sysName = \"IEEE14\"\n",
    "testType = \"ModelsFinal\"\n",
    "numEpochs = 100\n",
    "checkpoint_path = testType+\"/Saved Models/\"+sysName+\"_models/\"\n",
    "\n",
    "X1 = pd.read_csv(\"Data/\"+sysName+\"/\"+sysName+\"Data_40k_lowSparsity\"+\".csv\")\n",
    "y1 = pd.read_csv(\"Data/\"+sysName+\"/\"+sysName+\"Labels_40k_lowSparsity\"+\".csv\")\n",
    "X2 = pd.read_csv(\"Data/\"+sysName+\"/\"+sysName+\"Data_40k_midSparsity\"+\".csv\")\n",
    "y2 = pd.read_csv(\"Data/\"+sysName+\"/\"+sysName+\"Labels_40k_midSparsity\"+\".csv\")\n",
    "X3 = pd.read_csv(\"Data/\"+sysName+\"/\"+sysName+\"Data_40k_highSparsity\"+\".csv\")\n",
    "y3 = pd.read_csv(\"Data/\"+sysName+\"/\"+sysName+\"Labels_40k_highSparsity\"+\".csv\")\n",
    "\n",
    "Xv1 = pd.read_csv(\"Data/\"+sysName+\"/\"+sysName+\"Data_10k_lowSparsity\"+\".csv\")\n",
    "yv1 = pd.read_csv(\"Data/\"+sysName+\"/\"+sysName+\"Labels_10k_lowSparsity\"+\".csv\")\n",
    "Xv2 = pd.read_csv(\"Data/\"+sysName+\"/\"+sysName+\"Data_10k_midSparsity\"+\".csv\")\n",
    "yv2 = pd.read_csv(\"Data/\"+sysName+\"/\"+sysName+\"Labels_10k_midSparsity\"+\".csv\")\n",
    "Xv3 = pd.read_csv(\"Data/\"+sysName+\"/\"+sysName+\"Data_10k_highSparsity\"+\".csv\")\n",
    "yv3 = pd.read_csv(\"Data/\"+sysName+\"/\"+sysName+\"Labels_10k_highSparsity\"+\".csv\")\n",
    "\n",
    "Xt1 = pd.read_csv(\"Data/\"+sysName+\"/\"+sysName+\"Data_4k_lowSparsity\"+\".csv\")\n",
    "yt1 = pd.read_csv(\"Data/\"+sysName+\"/\"+sysName+\"Labels_4k_lowSparsity\"+\".csv\")\n",
    "Xt2 = pd.read_csv(\"Data/\"+sysName+\"/\"+sysName+\"Data_4k_midSparsity\"+\".csv\")\n",
    "yt2 = pd.read_csv(\"Data/\"+sysName+\"/\"+sysName+\"Labels_4k_midSparsity\"+\".csv\")\n",
    "Xt3 = pd.read_csv(\"Data/\"+sysName+\"/\"+sysName+\"Data_4k_highSparsity\"+\".csv\")\n",
    "yt3 = pd.read_csv(\"Data/\"+sysName+\"/\"+sysName+\"Labels_4k_highSparsity\"+\".csv\")\n",
    "\n",
    "X1,y1 = dataProc(X1,y1)\n",
    "Xv1,yv1 = dataProc(Xv1,yv1)\n",
    "Xt1,yt1 = dataProc(Xt1,yt1)\n",
    "X2,y2 = dataProc(X2,y2)\n",
    "Xv2,yv2 = dataProc(Xv2,yv2)\n",
    "Xt2,yt2 = dataProc(Xt2,yt2)\n",
    "\n",
    "X3,y3 = dataProc(X3,y3)\n",
    "Xv3,yv3 = dataProc(Xv3,yv3)\n",
    "Xt3,yt3 = dataProc(Xt3,yt3)\n",
    "\n",
    "numFeatures = len(X1[0])\n",
    "\n",
    "X = np.concatenate((X1, X2, X3), axis=0)\n",
    "y = np.concatenate((y1, y2, y3), axis=0)\n",
    "Xv = np.concatenate((Xv1, Xv2, Xv3), axis=0)\n",
    "yv = np.concatenate((yv1, yv2, yv3), axis=0)\n",
    "Xt = np.concatenate((Xt1, Xt2, Xt3), axis=0)\n",
    "yt = np.concatenate((yt1, yt2, yt3), axis=0)\n",
    "\n",
    "X, y = shuffle(X, y, random_state=0)\n",
    "Xv, yv = shuffle(Xv, yv, random_state=0)\n",
    "Xt, yt = shuffle(Xt, yt, random_state=0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "savedModels_path = 'Saved Models'\n",
    "\n",
    "resultsLow = []\n",
    "resultsMid = []\n",
    "resultsHigh = []"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Training"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "NUM_EPOCHS = 50"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "collapsed": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "2-layers-16-nodes\n",
      "Train on 119997 samples, validate on 29997 samples\n",
      "Epoch 1/50\n",
      "119997/119997 [==============================] - 5s 40us/sample - loss: 0.6902 - accuracy: 0.5147 - val_loss: 0.6789 - val_accuracy: 0.5422\n",
      "Epoch 2/50\n",
      "119997/119997 [==============================] - 4s 34us/sample - loss: 0.6801 - accuracy: 0.5504 - val_loss: 0.6826 - val_accuracy: 0.5021\n",
      "Epoch 3/50\n",
      "119997/119997 [==============================] - 4s 35us/sample - loss: 0.6578 - accuracy: 0.5832 - val_loss: 0.7166 - val_accuracy: 0.5121\n",
      "Epoch 4/50\n",
      "119997/119997 [==============================] - 4s 34us/sample - loss: 0.6416 - accuracy: 0.5947 - val_loss: 0.7585 - val_accuracy: 0.5080\n",
      "Epoch 5/50\n",
      "119997/119997 [==============================] - 4s 31us/sample - loss: 0.6332 - accuracy: 0.5976 - val_loss: 0.7929 - val_accuracy: 0.5130\n",
      "Epoch 6/50\n",
      "119997/119997 [==============================] - 4s 33us/sample - loss: 0.6283 - accuracy: 0.5995 - val_loss: 0.8540 - val_accuracy: 0.5156\n",
      "Epoch 7/50\n",
      "119997/119997 [==============================] - 4s 34us/sample - loss: 0.6244 - accuracy: 0.6016 - val_loss: 0.8883 - val_accuracy: 0.5834\n",
      "Epoch 8/50\n",
      "119997/119997 [==============================] - 4s 33us/sample - loss: 0.6198 - accuracy: 0.6065 - val_loss: 0.9726 - val_accuracy: 0.5698\n",
      "Epoch 9/50\n",
      "119997/119997 [==============================] - 4s 34us/sample - loss: 0.6102 - accuracy: 0.6338 - val_loss: 1.1164 - val_accuracy: 0.5671\n",
      "Epoch 10/50\n",
      "119997/119997 [==============================] - 4s 36us/sample - loss: 0.5949 - accuracy: 0.6687 - val_loss: 1.3264 - val_accuracy: 0.5870\n",
      "Epoch 11/50\n",
      "119997/119997 [==============================] - 4s 37us/sample - loss: 0.5811 - accuracy: 0.6918 - val_loss: 1.5299 - val_accuracy: 0.5876\n",
      "Epoch 12/50\n",
      "119997/119997 [==============================] - 4s 33us/sample - loss: 0.5698 - accuracy: 0.7039 - val_loss: 1.6580 - val_accuracy: 0.5937\n",
      "Epoch 13/50\n",
      "119997/119997 [==============================] - 4s 32us/sample - loss: 0.5530 - accuracy: 0.7249 - val_loss: 1.8803 - val_accuracy: 0.5897\n",
      "Epoch 14/50\n",
      "119997/119997 [==============================] - 4s 32us/sample - loss: 0.5419 - accuracy: 0.7298 - val_loss: 2.1238 - val_accuracy: 0.5988\n",
      "Epoch 15/50\n",
      "119997/119997 [==============================] - 4s 32us/sample - loss: 0.5347 - accuracy: 0.7287 - val_loss: 2.3375 - val_accuracy: 0.5818\n",
      "Epoch 16/50\n",
      "119997/119997 [==============================] - 4s 33us/sample - loss: 0.5270 - accuracy: 0.7324 - val_loss: 2.2180 - val_accuracy: 0.6020\n",
      "Epoch 17/50\n",
      "119997/119997 [==============================] - 4s 33us/sample - loss: 0.5221 - accuracy: 0.7353 - val_loss: 2.3739 - val_accuracy: 0.5886\n",
      "Epoch 18/50\n",
      "119997/119997 [==============================] - 4s 32us/sample - loss: 0.5152 - accuracy: 0.7412 - val_loss: 2.4028 - val_accuracy: 0.5938\n",
      "Epoch 19/50\n",
      "119997/119997 [==============================] - 4s 32us/sample - loss: 0.5116 - accuracy: 0.7421 - val_loss: 2.2981 - val_accuracy: 0.6089\n",
      "Epoch 20/50\n",
      "119997/119997 [==============================] - 4s 34us/sample - loss: 0.5070 - accuracy: 0.7475 - val_loss: 2.3278 - val_accuracy: 0.6110\n",
      "Epoch 21/50\n",
      "119997/119997 [==============================] - 4s 36us/sample - loss: 0.5022 - accuracy: 0.7504 - val_loss: 2.3570 - val_accuracy: 0.6002\n",
      "Epoch 22/50\n",
      "119997/119997 [==============================] - 4s 34us/sample - loss: 0.5014 - accuracy: 0.7507 - val_loss: 2.2021 - val_accuracy: 0.5852\n",
      "Epoch 23/50\n",
      "119997/119997 [==============================] - 4s 32us/sample - loss: 0.4979 - accuracy: 0.7528 - val_loss: 2.1381 - val_accuracy: 0.5970\n",
      "Epoch 24/50\n",
      "119997/119997 [==============================] - 4s 37us/sample - loss: 0.4928 - accuracy: 0.7536 - val_loss: 2.0535 - val_accuracy: 0.5471\n",
      "Epoch 25/50\n",
      "119997/119997 [==============================] - 4s 36us/sample - loss: 0.4901 - accuracy: 0.7575 - val_loss: 2.2226 - val_accuracy: 0.6041\n",
      "Epoch 26/50\n",
      "119997/119997 [==============================] - 4s 33us/sample - loss: 0.4859 - accuracy: 0.7659 - val_loss: 2.0352 - val_accuracy: 0.5898\n",
      "Epoch 27/50\n",
      "119997/119997 [==============================] - 4s 33us/sample - loss: 0.4829 - accuracy: 0.7667 - val_loss: 1.9857 - val_accuracy: 0.6035\n",
      "Epoch 28/50\n",
      "119997/119997 [==============================] - 4s 32us/sample - loss: 0.4813 - accuracy: 0.7664 - val_loss: 1.8998 - val_accuracy: 0.5585\n",
      "Epoch 29/50\n",
      "119997/119997 [==============================] - 5s 44us/sample - loss: 0.4782 - accuracy: 0.7683 - val_loss: 1.9158 - val_accuracy: 0.6128\n",
      "Epoch 30/50\n",
      "119997/119997 [==============================] - 5s 44us/sample - loss: 0.4775 - accuracy: 0.7691 - val_loss: 1.8687 - val_accuracy: 0.6166\n",
      "Epoch 31/50\n",
      "119997/119997 [==============================] - 5s 38us/sample - loss: 0.4727 - accuracy: 0.7726 - val_loss: 1.7392 - val_accuracy: 0.6144\n",
      "Epoch 32/50\n",
      "119997/119997 [==============================] - 5s 38us/sample - loss: 0.4681 - accuracy: 0.7776 - val_loss: 1.8658 - val_accuracy: 0.6321\n",
      "Epoch 33/50\n",
      "119997/119997 [==============================] - 5s 40us/sample - loss: 0.4667 - accuracy: 0.7777 - val_loss: 1.8191 - val_accuracy: 0.6147\n",
      "Epoch 34/50\n",
      "119997/119997 [==============================] - 5s 39us/sample - loss: 0.4632 - accuracy: 0.7797 - val_loss: 1.8651 - val_accuracy: 0.6090\n",
      "Epoch 35/50\n",
      "119997/119997 [==============================] - 5s 40us/sample - loss: 0.4614 - accuracy: 0.7793 - val_loss: 1.8866 - val_accuracy: 0.5711\n",
      "Epoch 36/50\n",
      "119997/119997 [==============================] - 5s 39us/sample - loss: 0.4578 - accuracy: 0.7827 - val_loss: 1.8243 - val_accuracy: 0.6448\n",
      "Epoch 37/50\n",
      "119997/119997 [==============================] - 5s 40us/sample - loss: 0.4546 - accuracy: 0.7851 - val_loss: 1.8917 - val_accuracy: 0.6395\n",
      "Epoch 38/50\n",
      "119997/119997 [==============================] - 4s 37us/sample - loss: 0.4510 - accuracy: 0.7880 - val_loss: 1.9357 - val_accuracy: 0.6540\n",
      "Epoch 39/50\n",
      "119997/119997 [==============================] - 4s 37us/sample - loss: 0.4500 - accuracy: 0.7893 - val_loss: 1.9885 - val_accuracy: 0.6387\n",
      "Epoch 40/50\n",
      "119997/119997 [==============================] - 5s 38us/sample - loss: 0.4513 - accuracy: 0.7866 - val_loss: 1.9158 - val_accuracy: 0.6325\n",
      "Epoch 41/50\n",
      "119997/119997 [==============================] - 4s 37us/sample - loss: 0.4442 - accuracy: 0.7931 - val_loss: 1.9871 - val_accuracy: 0.6244\n",
      "Epoch 42/50\n",
      "119997/119997 [==============================] - 4s 36us/sample - loss: 0.4419 - accuracy: 0.7937 - val_loss: 1.8636 - val_accuracy: 0.6118\n",
      "Epoch 43/50\n",
      "119997/119997 [==============================] - 4s 36us/sample - loss: 0.4412 - accuracy: 0.7926 - val_loss: 2.0976 - val_accuracy: 0.6102\n",
      "Epoch 44/50\n",
      "119997/119997 [==============================] - 4s 37us/sample - loss: 0.4328 - accuracy: 0.7994 - val_loss: 2.0921 - val_accuracy: 0.6341\n",
      "Epoch 45/50\n",
      "119997/119997 [==============================] - 4s 36us/sample - loss: 0.4371 - accuracy: 0.7952 - val_loss: 2.0836 - val_accuracy: 0.6216\n",
      "Epoch 46/50\n",
      "119997/119997 [==============================] - 4s 36us/sample - loss: 0.4294 - accuracy: 0.8029 - val_loss: 2.0013 - val_accuracy: 0.5906\n",
      "Epoch 47/50\n",
      "119997/119997 [==============================] - 4s 35us/sample - loss: 0.4359 - accuracy: 0.7974 - val_loss: 2.0446 - val_accuracy: 0.6055\n",
      "Epoch 48/50\n",
      "119997/119997 [==============================] - 5s 39us/sample - loss: 0.4330 - accuracy: 0.7973 - val_loss: 2.0643 - val_accuracy: 0.5610\n",
      "Epoch 49/50\n",
      "119997/119997 [==============================] - 4s 36us/sample - loss: 0.4300 - accuracy: 0.8010 - val_loss: 1.8362 - val_accuracy: 0.5858\n",
      "Epoch 50/50\n",
      "119997/119997 [==============================] - 5s 38us/sample - loss: 0.4281 - accuracy: 0.8022 - val_loss: 1.9021 - val_accuracy: 0.5967\n",
      "3999/3999 [==============================] - 0s 22us/sample - loss: 4.6253 - accuracy: 0.4569\n",
      "3999/3999 [==============================] - 0s 20us/sample - loss: 1.2466 - accuracy: 0.5531\n",
      "3999/3999 [==============================] - 0s 19us/sample - loss: 17.2572 - accuracy: 0.5974\n",
      "2-layers-32-nodes\n",
      "Train on 119997 samples, validate on 29997 samples\n",
      "Epoch 1/50\n",
      "119997/119997 [==============================] - 5s 42us/sample - loss: 0.6921 - accuracy: 0.5109 - val_loss: 0.6716 - val_accuracy: 0.5223\n",
      "Epoch 2/50\n",
      "119997/119997 [==============================] - 5s 40us/sample - loss: 0.6880 - accuracy: 0.5183 - val_loss: 0.6633 - val_accuracy: 0.5912\n",
      "Epoch 3/50\n",
      "119997/119997 [==============================] - 5s 40us/sample - loss: 0.6846 - accuracy: 0.5278 - val_loss: 0.6778 - val_accuracy: 0.5269\n",
      "Epoch 4/50\n",
      "119997/119997 [==============================] - 4s 36us/sample - loss: 0.6788 - accuracy: 0.5462 - val_loss: 0.8165 - val_accuracy: 0.4920\n",
      "Epoch 5/50\n",
      "119997/119997 [==============================] - 4s 35us/sample - loss: 0.6626 - accuracy: 0.5795 - val_loss: 0.9467 - val_accuracy: 0.5332\n",
      "Epoch 6/50\n",
      "119997/119997 [==============================] - 4s 35us/sample - loss: 0.6396 - accuracy: 0.6032 - val_loss: 1.0832 - val_accuracy: 0.5396\n",
      "Epoch 7/50\n",
      "119997/119997 [==============================] - 4s 34us/sample - loss: 0.6143 - accuracy: 0.6428 - val_loss: 1.2667 - val_accuracy: 0.5506\n",
      "Epoch 8/50\n",
      "119997/119997 [==============================] - 4s 34us/sample - loss: 0.5870 - accuracy: 0.6788 - val_loss: 1.6590 - val_accuracy: 0.5410\n",
      "Epoch 9/50\n",
      "119997/119997 [==============================] - 4s 34us/sample - loss: 0.5641 - accuracy: 0.6909 - val_loss: 2.1254 - val_accuracy: 0.5690\n",
      "Epoch 10/50\n",
      "119997/119997 [==============================] - 4s 32us/sample - loss: 0.5487 - accuracy: 0.7000 - val_loss: 2.2701 - val_accuracy: 0.5817\n",
      "Epoch 11/50\n",
      "119997/119997 [==============================] - 4s 32us/sample - loss: 0.5369 - accuracy: 0.7144 - val_loss: 2.8147 - val_accuracy: 0.5914\n",
      "Epoch 12/50\n",
      "119997/119997 [==============================] - 4s 32us/sample - loss: 0.5291 - accuracy: 0.7211 - val_loss: 3.0934 - val_accuracy: 0.5947\n",
      "Epoch 13/50\n",
      "119997/119997 [==============================] - 4s 32us/sample - loss: 0.5224 - accuracy: 0.7273 - val_loss: 3.5679 - val_accuracy: 0.5998\n",
      "Epoch 14/50\n",
      "119997/119997 [==============================] - 4s 32us/sample - loss: 0.5181 - accuracy: 0.7341 - val_loss: 3.8274 - val_accuracy: 0.6015\n",
      "Epoch 15/50\n",
      "119997/119997 [==============================] - 4s 32us/sample - loss: 0.5125 - accuracy: 0.7361 - val_loss: 4.4633 - val_accuracy: 0.5862\n",
      "Epoch 16/50\n",
      "119997/119997 [==============================] - 4s 32us/sample - loss: 0.5064 - accuracy: 0.7418 - val_loss: 4.7237 - val_accuracy: 0.5865\n",
      "Epoch 17/50\n",
      "119997/119997 [==============================] - 4s 32us/sample - loss: 0.5011 - accuracy: 0.7461 - val_loss: 5.3478 - val_accuracy: 0.5745\n",
      "Epoch 18/50\n",
      "119997/119997 [==============================] - 4s 33us/sample - loss: 0.4941 - accuracy: 0.7496 - val_loss: 5.7208 - val_accuracy: 0.5752\n",
      "Epoch 19/50\n",
      "119997/119997 [==============================] - 4s 35us/sample - loss: 0.4932 - accuracy: 0.7506 - val_loss: 5.7119 - val_accuracy: 0.5795\n",
      "Epoch 20/50\n",
      "119997/119997 [==============================] - 4s 34us/sample - loss: 0.4897 - accuracy: 0.7540 - val_loss: 5.9496 - val_accuracy: 0.5889\n",
      "Epoch 21/50\n",
      "119997/119997 [==============================] - 4s 34us/sample - loss: 0.4878 - accuracy: 0.7563 - val_loss: 5.9817 - val_accuracy: 0.6022\n",
      "Epoch 22/50\n",
      "119997/119997 [==============================] - 4s 36us/sample - loss: 0.4830 - accuracy: 0.7590 - val_loss: 6.6945 - val_accuracy: 0.5789\n",
      "Epoch 23/50\n",
      "119997/119997 [==============================] - 4s 33us/sample - loss: 0.4812 - accuracy: 0.7619 - val_loss: 7.8806 - val_accuracy: 0.5684\n",
      "Epoch 24/50\n",
      "119997/119997 [==============================] - 4s 33us/sample - loss: 0.4772 - accuracy: 0.7664 - val_loss: 7.3684 - val_accuracy: 0.5741\n",
      "Epoch 25/50\n",
      "119997/119997 [==============================] - 4s 32us/sample - loss: 0.4767 - accuracy: 0.7690 - val_loss: 8.4899 - val_accuracy: 0.5687\n",
      "Epoch 26/50\n",
      "119997/119997 [==============================] - 4s 35us/sample - loss: 0.4732 - accuracy: 0.7675 - val_loss: 8.9906 - val_accuracy: 0.5554\n",
      "Epoch 27/50\n",
      "119997/119997 [==============================] - 4s 37us/sample - loss: 0.4713 - accuracy: 0.7713 - val_loss: 9.6081 - val_accuracy: 0.5610\n",
      "Epoch 28/50\n",
      "119997/119997 [==============================] - 4s 33us/sample - loss: 0.4688 - accuracy: 0.7756 - val_loss: 7.7435 - val_accuracy: 0.6023\n",
      "Epoch 29/50\n",
      "119997/119997 [==============================] - 4s 33us/sample - loss: 0.4652 - accuracy: 0.7761 - val_loss: 9.4882 - val_accuracy: 0.5663\n",
      "Epoch 30/50\n",
      "119997/119997 [==============================] - 4s 34us/sample - loss: 0.4653 - accuracy: 0.7772 - val_loss: 8.8769 - val_accuracy: 0.5750\n",
      "Epoch 31/50\n",
      "119997/119997 [==============================] - 4s 31us/sample - loss: 0.4603 - accuracy: 0.7799 - val_loss: 9.9631 - val_accuracy: 0.5681\n",
      "Epoch 32/50\n",
      "119997/119997 [==============================] - 4s 31us/sample - loss: 0.4637 - accuracy: 0.7781 - val_loss: 8.9240 - val_accuracy: 0.6018\n",
      "Epoch 33/50\n",
      "119997/119997 [==============================] - 4s 32us/sample - loss: 0.4600 - accuracy: 0.7805 - val_loss: 8.6583 - val_accuracy: 0.6043\n",
      "Epoch 34/50\n",
      "119997/119997 [==============================] - 4s 32us/sample - loss: 0.4564 - accuracy: 0.7842 - val_loss: 9.4022 - val_accuracy: 0.6023\n",
      "Epoch 35/50\n",
      "119997/119997 [==============================] - 4s 32us/sample - loss: 0.4529 - accuracy: 0.7885 - val_loss: 9.3822 - val_accuracy: 0.6098\n",
      "Epoch 36/50\n",
      "119997/119997 [==============================] - 4s 31us/sample - loss: 0.4517 - accuracy: 0.7861 - val_loss: 12.4958 - val_accuracy: 0.5533\n",
      "Epoch 37/50\n",
      "119997/119997 [==============================] - 4s 32us/sample - loss: 0.4482 - accuracy: 0.7908 - val_loss: 11.4386 - val_accuracy: 0.5531\n",
      "Epoch 38/50\n",
      "119997/119997 [==============================] - 4s 36us/sample - loss: 0.4490 - accuracy: 0.7904 - val_loss: 12.5371 - val_accuracy: 0.5437\n",
      "Epoch 39/50\n",
      "119997/119997 [==============================] - 4s 34us/sample - loss: 0.4474 - accuracy: 0.7933 - val_loss: 13.7535 - val_accuracy: 0.5448\n",
      "Epoch 40/50\n",
      "119997/119997 [==============================] - 5s 38us/sample - loss: 0.4453 - accuracy: 0.7910 - val_loss: 10.5847 - val_accuracy: 0.5683\n",
      "Epoch 41/50\n",
      "119997/119997 [==============================] - 4s 35us/sample - loss: 0.4406 - accuracy: 0.7959 - val_loss: 12.6499 - val_accuracy: 0.5565\n",
      "Epoch 42/50\n",
      "119997/119997 [==============================] - 4s 33us/sample - loss: 0.4387 - accuracy: 0.7992 - val_loss: 14.4076 - val_accuracy: 0.5449\n",
      "Epoch 43/50\n",
      "119997/119997 [==============================] - 4s 32us/sample - loss: 0.4332 - accuracy: 0.8030 - val_loss: 10.7580 - val_accuracy: 0.5678\n",
      "Epoch 44/50\n",
      "119997/119997 [==============================] - 4s 32us/sample - loss: 0.4305 - accuracy: 0.8046 - val_loss: 14.1489 - val_accuracy: 0.5491\n",
      "Epoch 45/50\n",
      "119997/119997 [==============================] - 4s 33us/sample - loss: 0.4280 - accuracy: 0.8053 - val_loss: 13.9607 - val_accuracy: 0.5495\n",
      "Epoch 46/50\n",
      "119997/119997 [==============================] - 4s 32us/sample - loss: 0.4221 - accuracy: 0.8106 - val_loss: 15.6330 - val_accuracy: 0.5445\n",
      "Epoch 47/50\n",
      "119997/119997 [==============================] - 4s 32us/sample - loss: 0.4190 - accuracy: 0.8136 - val_loss: 12.5883 - val_accuracy: 0.5190\n",
      "Epoch 48/50\n",
      "119997/119997 [==============================] - 4s 32us/sample - loss: 0.4156 - accuracy: 0.8149 - val_loss: 11.7448 - val_accuracy: 0.5819\n",
      "Epoch 49/50\n",
      "119997/119997 [==============================] - 4s 32us/sample - loss: 0.4115 - accuracy: 0.8190 - val_loss: 13.6990 - val_accuracy: 0.5477\n",
      "Epoch 50/50\n",
      "119997/119997 [==============================] - 4s 32us/sample - loss: 0.4088 - accuracy: 0.8199 - val_loss: 14.4303 - val_accuracy: 0.5349\n",
      "3999/3999 [==============================] - 0s 16us/sample - loss: 22.2402 - accuracy: 0.4806\n",
      "3999/3999 [==============================] - 0s 21us/sample - loss: 4.1687 - accuracy: 0.7029\n",
      "3999/3999 [==============================] - 0s 17us/sample - loss: 23.2406 - accuracy: 0.5786\n",
      "2-layers-64-nodes\n",
      "Train on 119997 samples, validate on 29997 samples\n",
      "Epoch 1/50\n",
      "119997/119997 [==============================] - 5s 39us/sample - loss: 0.6928 - accuracy: 0.5091 - val_loss: 0.6840 - val_accuracy: 0.4826\n",
      "Epoch 2/50\n",
      "119997/119997 [==============================] - 4s 36us/sample - loss: 0.6883 - accuracy: 0.5184 - val_loss: 0.6867 - val_accuracy: 0.5143\n",
      "Epoch 3/50\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "119997/119997 [==============================] - 5s 38us/sample - loss: 0.6849 - accuracy: 0.5270 - val_loss: 0.7281 - val_accuracy: 0.4905\n",
      "Epoch 4/50\n",
      "119997/119997 [==============================] - 4s 34us/sample - loss: 0.6792 - accuracy: 0.5406 - val_loss: 0.7670 - val_accuracy: 0.4869\n",
      "Epoch 5/50\n",
      "119997/119997 [==============================] - 4s 33us/sample - loss: 0.6580 - accuracy: 0.5834 - val_loss: 0.8395 - val_accuracy: 0.4902\n",
      "Epoch 6/50\n",
      "119997/119997 [==============================] - 4s 34us/sample - loss: 0.6270 - accuracy: 0.6268 - val_loss: 0.9477 - val_accuracy: 0.5344\n",
      "Epoch 7/50\n",
      "119997/119997 [==============================] - 4s 33us/sample - loss: 0.5898 - accuracy: 0.6734 - val_loss: 1.1157 - val_accuracy: 0.5330\n",
      "Epoch 8/50\n",
      "119997/119997 [==============================] - 4s 33us/sample - loss: 0.5606 - accuracy: 0.6869 - val_loss: 1.4135 - val_accuracy: 0.5336\n",
      "Epoch 9/50\n",
      "119997/119997 [==============================] - 4s 33us/sample - loss: 0.5462 - accuracy: 0.6937 - val_loss: 1.4807 - val_accuracy: 0.54785473 - \n",
      "Epoch 10/50\n",
      "119997/119997 [==============================] - 4s 33us/sample - loss: 0.5364 - accuracy: 0.7044 - val_loss: 1.8255 - val_accuracy: 0.5378\n",
      "Epoch 11/50\n",
      "119997/119997 [==============================] - 4s 34us/sample - loss: 0.5275 - accuracy: 0.7159 - val_loss: 2.1077 - val_accuracy: 0.5479\n",
      "Epoch 12/50\n",
      "119997/119997 [==============================] - 4s 35us/sample - loss: 0.5180 - accuracy: 0.7309 - val_loss: 2.0771 - val_accuracy: 0.5429\n",
      "Epoch 13/50\n",
      "119997/119997 [==============================] - 4s 34us/sample - loss: 0.5082 - accuracy: 0.7414 - val_loss: 2.3387 - val_accuracy: 0.5485\n",
      "Epoch 14/50\n",
      "119997/119997 [==============================] - 4s 33us/sample - loss: 0.4976 - accuracy: 0.7478 - val_loss: 2.4267 - val_accuracy: 0.5540\n",
      "Epoch 15/50\n",
      "119997/119997 [==============================] - 4s 33us/sample - loss: 0.4915 - accuracy: 0.7527 - val_loss: 3.0337 - val_accuracy: 0.5586\n",
      "Epoch 16/50\n",
      "119997/119997 [==============================] - 4s 33us/sample - loss: 0.4842 - accuracy: 0.7577 - val_loss: 2.6504 - val_accuracy: 0.5552\n",
      "Epoch 17/50\n",
      "119997/119997 [==============================] - 4s 33us/sample - loss: 0.4785 - accuracy: 0.7643 - val_loss: 3.0753 - val_accuracy: 0.5607\n",
      "Epoch 18/50\n",
      "119997/119997 [==============================] - 4s 33us/sample - loss: 0.4748 - accuracy: 0.7679 - val_loss: 3.3279 - val_accuracy: 0.5647\n",
      "Epoch 19/50\n",
      "119997/119997 [==============================] - 4s 32us/sample - loss: 0.4708 - accuracy: 0.7684 - val_loss: 3.3522 - val_accuracy: 0.5731\n",
      "Epoch 20/50\n",
      "119997/119997 [==============================] - 4s 33us/sample - loss: 0.4665 - accuracy: 0.7725 - val_loss: 3.1837 - val_accuracy: 0.5687\n",
      "Epoch 21/50\n",
      "119997/119997 [==============================] - 4s 33us/sample - loss: 0.4617 - accuracy: 0.7765 - val_loss: 3.1854 - val_accuracy: 0.5712\n",
      "Epoch 22/50\n",
      "119997/119997 [==============================] - 4s 33us/sample - loss: 0.4526 - accuracy: 0.7856 - val_loss: 3.3994 - val_accuracy: 0.5772\n",
      "Epoch 23/50\n",
      "119997/119997 [==============================] - 4s 33us/sample - loss: 0.4513 - accuracy: 0.7854 - val_loss: 3.6403 - val_accuracy: 0.5767\n",
      "Epoch 24/50\n",
      "119997/119997 [==============================] - 4s 33us/sample - loss: 0.4477 - accuracy: 0.7893 - val_loss: 4.1999 - val_accuracy: 0.5828\n",
      "Epoch 25/50\n",
      "119997/119997 [==============================] - 4s 33us/sample - loss: 0.4477 - accuracy: 0.7868 - val_loss: 3.8428 - val_accuracy: 0.5781\n",
      "Epoch 26/50\n",
      "119997/119997 [==============================] - 4s 33us/sample - loss: 0.4375 - accuracy: 0.7946 - val_loss: 4.0264 - val_accuracy: 0.5836\n",
      "Epoch 27/50\n",
      "119997/119997 [==============================] - 4s 32us/sample - loss: 0.4381 - accuracy: 0.7939 - val_loss: 3.9964 - val_accuracy: 0.5799\n",
      "Epoch 28/50\n",
      "119997/119997 [==============================] - 4s 32us/sample - loss: 0.4322 - accuracy: 0.7964 - val_loss: 4.1730 - val_accuracy: 0.5785\n",
      "Epoch 29/50\n",
      "119997/119997 [==============================] - 4s 32us/sample - loss: 0.4283 - accuracy: 0.7996 - val_loss: 4.2134 - val_accuracy: 0.5756\n",
      "Epoch 30/50\n",
      "119997/119997 [==============================] - 4s 32us/sample - loss: 0.4267 - accuracy: 0.8011 - val_loss: 4.5619 - val_accuracy: 0.5941\n",
      "Epoch 31/50\n",
      "119997/119997 [==============================] - 4s 33us/sample - loss: 0.4234 - accuracy: 0.8038 - val_loss: 4.5940 - val_accuracy: 0.5851\n",
      "Epoch 32/50\n",
      "119997/119997 [==============================] - 4s 33us/sample - loss: 0.4210 - accuracy: 0.8041 - val_loss: 4.6303 - val_accuracy: 0.5831\n",
      "Epoch 33/50\n",
      "119997/119997 [==============================] - 4s 33us/sample - loss: 0.4155 - accuracy: 0.8085 - val_loss: 4.7208 - val_accuracy: 0.5769\n",
      "Epoch 34/50\n",
      "119997/119997 [==============================] - 4s 33us/sample - loss: 0.4186 - accuracy: 0.8065 - val_loss: 4.7718 - val_accuracy: 0.5821\n",
      "Epoch 35/50\n",
      "119997/119997 [==============================] - 4s 33us/sample - loss: 0.4125 - accuracy: 0.8099 - val_loss: 4.8957 - val_accuracy: 0.5833\n",
      "Epoch 36/50\n",
      "119997/119997 [==============================] - 4s 33us/sample - loss: 0.4123 - accuracy: 0.8104 - val_loss: 5.3907 - val_accuracy: 0.5883\n",
      "Epoch 37/50\n",
      "119997/119997 [==============================] - 4s 34us/sample - loss: 0.4090 - accuracy: 0.8105 - val_loss: 5.6782 - val_accuracy: 0.5869\n",
      "Epoch 38/50\n",
      "119997/119997 [==============================] - 4s 33us/sample - loss: 0.4043 - accuracy: 0.8151 - val_loss: 5.5230 - val_accuracy: 0.5936\n",
      "Epoch 39/50\n",
      "119997/119997 [==============================] - 4s 32us/sample - loss: 0.4097 - accuracy: 0.8113 - val_loss: 6.1940 - val_accuracy: 0.5860\n",
      "Epoch 40/50\n",
      "119997/119997 [==============================] - 4s 33us/sample - loss: 0.4074 - accuracy: 0.8168 - val_loss: 5.5896 - val_accuracy: 0.5878\n",
      "Epoch 41/50\n",
      "119997/119997 [==============================] - 4s 33us/sample - loss: 0.4109 - accuracy: 0.8141 - val_loss: 4.4356 - val_accuracy: 0.5868\n",
      "Epoch 42/50\n",
      "119997/119997 [==============================] - 4s 33us/sample - loss: 0.4092 - accuracy: 0.8148 - val_loss: 3.7782 - val_accuracy: 0.5906\n",
      "Epoch 43/50\n",
      "119997/119997 [==============================] - 4s 32us/sample - loss: 0.4051 - accuracy: 0.8164 - val_loss: 3.4698 - val_accuracy: 0.6018\n",
      "Epoch 44/50\n",
      "119997/119997 [==============================] - 4s 32us/sample - loss: 0.4054 - accuracy: 0.8161 - val_loss: 3.5541 - val_accuracy: 0.5992\n",
      "Epoch 45/50\n",
      "119997/119997 [==============================] - 4s 33us/sample - loss: 0.4023 - accuracy: 0.8189 - val_loss: 3.7554 - val_accuracy: 0.5971\n",
      "Epoch 46/50\n",
      "119997/119997 [==============================] - 4s 33us/sample - loss: 0.3979 - accuracy: 0.8228 - val_loss: 3.4752 - val_accuracy: 0.5739\n",
      "Epoch 47/50\n",
      "119997/119997 [==============================] - 4s 33us/sample - loss: 0.3996 - accuracy: 0.8192 - val_loss: 4.1737 - val_accuracy: 0.5472\n",
      "Epoch 48/50\n",
      "119997/119997 [==============================] - 4s 33us/sample - loss: 0.3994 - accuracy: 0.8201 - val_loss: 2.9286 - val_accuracy: 0.5748\n",
      "Epoch 49/50\n",
      "119997/119997 [==============================] - 4s 32us/sample - loss: 0.4012 - accuracy: 0.8157 - val_loss: 3.9104 - val_accuracy: 0.6020\n",
      "Epoch 50/50\n",
      "119997/119997 [==============================] - 4s 34us/sample - loss: 0.3910 - accuracy: 0.8272 - val_loss: 4.1501 - val_accuracy: 0.5983\n",
      "3999/3999 [==============================] - 0s 25us/sample - loss: 8.6609 - accuracy: 0.4889\n",
      "3999/3999 [==============================] - 0s 21us/sample - loss: 0.5440 - accuracy: 0.5799\n",
      "3999/3999 [==============================] - 0s 20us/sample - loss: 10.6267 - accuracy: 0.6252\n",
      "2-layers-128-nodes\n",
      "Train on 119997 samples, validate on 29997 samples\n",
      "Epoch 1/50\n",
      "119997/119997 [==============================] - 5s 43us/sample - loss: 0.6929 - accuracy: 0.5102 - val_loss: 0.6709 - val_accuracy: 0.5711\n",
      "Epoch 2/50\n",
      "119997/119997 [==============================] - 4s 33us/sample - loss: 0.6884 - accuracy: 0.5175 - val_loss: 0.6596 - val_accuracy: 0.5235\n",
      "Epoch 3/50\n",
      "119997/119997 [==============================] - 4s 33us/sample - loss: 0.6847 - accuracy: 0.5285 - val_loss: 0.6599 - val_accuracy: 0.5653\n",
      "Epoch 4/50\n",
      "119997/119997 [==============================] - 4s 37us/sample - loss: 0.6763 - accuracy: 0.5556 - val_loss: 0.7328 - val_accuracy: 0.5374\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 5/50\n",
      "119997/119997 [==============================] - 5s 41us/sample - loss: 0.6475 - accuracy: 0.6135 - val_loss: 0.9281 - val_accuracy: 0.5422\n",
      "Epoch 6/50\n",
      "119997/119997 [==============================] - 4s 37us/sample - loss: 0.6036 - accuracy: 0.6640 - val_loss: 1.1884 - val_accuracy: 0.5725\n",
      "Epoch 7/50\n",
      "119997/119997 [==============================] - 4s 34us/sample - loss: 0.5764 - accuracy: 0.6724 - val_loss: 1.2503 - val_accuracy: 0.6022\n",
      "Epoch 8/50\n",
      "119997/119997 [==============================] - 4s 34us/sample - loss: 0.5605 - accuracy: 0.6861 - val_loss: 1.4633 - val_accuracy: 0.5943\n",
      "Epoch 9/50\n",
      "119997/119997 [==============================] - 4s 33us/sample - loss: 0.5468 - accuracy: 0.7057 - val_loss: 1.8453 - val_accuracy: 0.5183\n",
      "Epoch 10/50\n",
      "119997/119997 [==============================] - 4s 34us/sample - loss: 0.5359 - accuracy: 0.7158 - val_loss: 1.6460 - val_accuracy: 0.6083\n",
      "Epoch 11/50\n",
      "119997/119997 [==============================] - 4s 34us/sample - loss: 0.5255 - accuracy: 0.7272 - val_loss: 1.8657 - val_accuracy: 0.6009\n",
      "Epoch 12/50\n",
      "119997/119997 [==============================] - 4s 34us/sample - loss: 0.5178 - accuracy: 0.7342 - val_loss: 2.2921 - val_accuracy: 0.6012\n",
      "Epoch 13/50\n",
      "119997/119997 [==============================] - 4s 34us/sample - loss: 0.5090 - accuracy: 0.7379 - val_loss: 1.9031 - val_accuracy: 0.5658\n",
      "Epoch 14/50\n",
      "119997/119997 [==============================] - 4s 34us/sample - loss: 0.5076 - accuracy: 0.7362 - val_loss: 2.1451 - val_accuracy: 0.5685\n",
      "Epoch 15/50\n",
      "119997/119997 [==============================] - 4s 34us/sample - loss: 0.5002 - accuracy: 0.7438 - val_loss: 2.1212 - val_accuracy: 0.6152\n",
      "Epoch 16/50\n",
      "119997/119997 [==============================] - 4s 35us/sample - loss: 0.4951 - accuracy: 0.7471 - val_loss: 2.3368 - val_accuracy: 0.6171\n",
      "Epoch 17/50\n",
      "119997/119997 [==============================] - 4s 34us/sample - loss: 0.4931 - accuracy: 0.7493 - val_loss: 2.6321 - val_accuracy: 0.5898\n",
      "Epoch 18/50\n",
      "119997/119997 [==============================] - 4s 34us/sample - loss: 0.4868 - accuracy: 0.7536 - val_loss: 2.5242 - val_accuracy: 0.5756\n",
      "Epoch 19/50\n",
      "119997/119997 [==============================] - 4s 36us/sample - loss: 0.4829 - accuracy: 0.7575 - val_loss: 2.6877 - val_accuracy: 0.6073\n",
      "Epoch 20/50\n",
      "119997/119997 [==============================] - 4s 33us/sample - loss: 0.4808 - accuracy: 0.7599 - val_loss: 2.7065 - val_accuracy: 0.6287\n",
      "Epoch 21/50\n",
      "119997/119997 [==============================] - 4s 34us/sample - loss: 0.4800 - accuracy: 0.7565 - val_loss: 2.9798 - val_accuracy: 0.6238\n",
      "Epoch 22/50\n",
      "119997/119997 [==============================] - 4s 34us/sample - loss: 0.4733 - accuracy: 0.7668 - val_loss: 3.1495 - val_accuracy: 0.6085\n",
      "Epoch 23/50\n",
      "119997/119997 [==============================] - 4s 33us/sample - loss: 0.4705 - accuracy: 0.7687 - val_loss: 3.1267 - val_accuracy: 0.5766\n",
      "Epoch 24/50\n",
      "119997/119997 [==============================] - 4s 34us/sample - loss: 0.4690 - accuracy: 0.7696 - val_loss: 3.0800 - val_accuracy: 0.5857\n",
      "Epoch 25/50\n",
      "119997/119997 [==============================] - 4s 34us/sample - loss: 0.4620 - accuracy: 0.7755 - val_loss: 3.4468 - val_accuracy: 0.5675\n",
      "Epoch 26/50\n",
      "119997/119997 [==============================] - 4s 34us/sample - loss: 0.4602 - accuracy: 0.7756 - val_loss: 3.3634 - val_accuracy: 0.5737\n",
      "Epoch 27/50\n",
      "119997/119997 [==============================] - 4s 35us/sample - loss: 0.4596 - accuracy: 0.7769 - val_loss: 3.3612 - val_accuracy: 0.6084\n",
      "Epoch 28/50\n",
      "119997/119997 [==============================] - 4s 34us/sample - loss: 0.4537 - accuracy: 0.7796 - val_loss: 3.5577 - val_accuracy: 0.5847\n",
      "Epoch 29/50\n",
      "119997/119997 [==============================] - 4s 34us/sample - loss: 0.4561 - accuracy: 0.7782 - val_loss: 3.7589 - val_accuracy: 0.5798\n",
      "Epoch 30/50\n",
      "119997/119997 [==============================] - 4s 34us/sample - loss: 0.4560 - accuracy: 0.7778 - val_loss: 3.7273 - val_accuracy: 0.6001\n",
      "Epoch 31/50\n",
      "119997/119997 [==============================] - 4s 34us/sample - loss: 0.4518 - accuracy: 0.7827 - val_loss: 4.1122 - val_accuracy: 0.5647\n",
      "Epoch 32/50\n",
      "119997/119997 [==============================] - 4s 35us/sample - loss: 0.4483 - accuracy: 0.7840 - val_loss: 4.0947 - val_accuracy: 0.5707\n",
      "Epoch 33/50\n",
      "119997/119997 [==============================] - 4s 34us/sample - loss: 0.4451 - accuracy: 0.7873 - val_loss: 3.9483 - val_accuracy: 0.5738\n",
      "Epoch 34/50\n",
      "119997/119997 [==============================] - 4s 34us/sample - loss: 0.4424 - accuracy: 0.7897 - val_loss: 4.0819 - val_accuracy: 0.6321\n",
      "Epoch 35/50\n",
      "119997/119997 [==============================] - 4s 34us/sample - loss: 0.4374 - accuracy: 0.7918 - val_loss: 4.1211 - val_accuracy: 0.5798\n",
      "Epoch 36/50\n",
      "119997/119997 [==============================] - 4s 33us/sample - loss: 0.4355 - accuracy: 0.7944 - val_loss: 4.6311 - val_accuracy: 0.5656\n",
      "Epoch 37/50\n",
      "119997/119997 [==============================] - 4s 34us/sample - loss: 0.4334 - accuracy: 0.7959 - val_loss: 4.5727 - val_accuracy: 0.5687\n",
      "Epoch 38/50\n",
      "119997/119997 [==============================] - 4s 33us/sample - loss: 0.4310 - accuracy: 0.7953 - val_loss: 4.3428 - val_accuracy: 0.5894\n",
      "Epoch 39/50\n",
      "119997/119997 [==============================] - 4s 34us/sample - loss: 0.4247 - accuracy: 0.8009 - val_loss: 4.5816 - val_accuracy: 0.5773\n",
      "Epoch 40/50\n",
      "119997/119997 [==============================] - 4s 34us/sample - loss: 0.4233 - accuracy: 0.8013 - val_loss: 4.7782 - val_accuracy: 0.5743\n",
      "Epoch 41/50\n",
      "119997/119997 [==============================] - 4s 35us/sample - loss: 0.4192 - accuracy: 0.8056 - val_loss: 5.0193 - val_accuracy: 0.5716\n",
      "Epoch 42/50\n",
      "119997/119997 [==============================] - 4s 34us/sample - loss: 0.4205 - accuracy: 0.8035 - val_loss: 5.1263 - val_accuracy: 0.5651\n",
      "Epoch 43/50\n",
      "119997/119997 [==============================] - 4s 34us/sample - loss: 0.4173 - accuracy: 0.8064 - val_loss: 4.9972 - val_accuracy: 0.5804\n",
      "Epoch 44/50\n",
      "119997/119997 [==============================] - 4s 34us/sample - loss: 0.4111 - accuracy: 0.8109 - val_loss: 5.6020 - val_accuracy: 0.5592\n",
      "Epoch 45/50\n",
      "119997/119997 [==============================] - 4s 37us/sample - loss: 0.4139 - accuracy: 0.8053 - val_loss: 5.1742 - val_accuracy: 0.5867\n",
      "Epoch 46/50\n",
      "119997/119997 [==============================] - 4s 33us/sample - loss: 0.4109 - accuracy: 0.8093 - val_loss: 5.4646 - val_accuracy: 0.5767\n",
      "Epoch 47/50\n",
      "119997/119997 [==============================] - 4s 34us/sample - loss: 0.4063 - accuracy: 0.8118 - val_loss: 6.0444 - val_accuracy: 0.5557\n",
      "Epoch 48/50\n",
      "119997/119997 [==============================] - 4s 34us/sample - loss: 0.4064 - accuracy: 0.8129 - val_loss: 5.3623 - val_accuracy: 0.5873\n",
      "Epoch 49/50\n",
      "119997/119997 [==============================] - 4s 34us/sample - loss: 0.4020 - accuracy: 0.8151 - val_loss: 6.1372 - val_accuracy: 0.5536\n",
      "Epoch 50/50\n",
      "119997/119997 [==============================] - 4s 34us/sample - loss: 0.3969 - accuracy: 0.8190 - val_loss: 6.0349 - val_accuracy: 0.5598\n",
      "3999/3999 [==============================] - 0s 16us/sample - loss: 13.0781 - accuracy: 0.4189\n",
      "3999/3999 [==============================] - 0s 17us/sample - loss: 1.8635 - accuracy: 0.5006\n",
      "3999/3999 [==============================] - 0s 16us/sample - loss: 12.6203 - accuracy: 0.6074\n",
      "3-layers-16-nodes\n",
      "Train on 119997 samples, validate on 29997 samples\n",
      "Epoch 1/50\n",
      "119997/119997 [==============================] - 5s 38us/sample - loss: 0.6909 - accuracy: 0.5081 - val_loss: 0.6595 - val_accuracy: 0.5589\n",
      "Epoch 2/50\n",
      "119997/119997 [==============================] - 4s 35us/sample - loss: 0.6863 - accuracy: 0.5204 - val_loss: 0.6765 - val_accuracy: 0.5358\n",
      "Epoch 3/50\n",
      "119997/119997 [==============================] - 4s 34us/sample - loss: 0.6802 - accuracy: 0.5360 - val_loss: 0.6809 - val_accuracy: 0.5736\n",
      "Epoch 4/50\n",
      "119997/119997 [==============================] - 4s 34us/sample - loss: 0.6547 - accuracy: 0.5755 - val_loss: 0.7769 - val_accuracy: 0.5336\n",
      "Epoch 5/50\n",
      "119997/119997 [==============================] - 4s 35us/sample - loss: 0.6051 - accuracy: 0.6463 - val_loss: 1.2469 - val_accuracy: 0.5455\n",
      "Epoch 6/50\n",
      "119997/119997 [==============================] - 4s 36us/sample - loss: 0.5717 - accuracy: 0.6614 - val_loss: 1.9594 - val_accuracy: 0.5597\n",
      "Epoch 7/50\n",
      "119997/119997 [==============================] - 4s 35us/sample - loss: 0.5599 - accuracy: 0.6657 - val_loss: 2.5757 - val_accuracy: 0.5564\n",
      "Epoch 8/50\n",
      "119997/119997 [==============================] - 4s 34us/sample - loss: 0.5543 - accuracy: 0.6676 - val_loss: 2.9353 - val_accuracy: 0.5590\n",
      "Epoch 9/50\n",
      "119997/119997 [==============================] - 4s 35us/sample - loss: 0.5479 - accuracy: 0.6719 - val_loss: 3.6055 - val_accuracy: 0.5629\n",
      "Epoch 10/50\n",
      "119997/119997 [==============================] - 4s 35us/sample - loss: 0.5472 - accuracy: 0.6732 - val_loss: 3.8537 - val_accuracy: 0.5729\n",
      "Epoch 11/50\n",
      "119997/119997 [==============================] - 4s 36us/sample - loss: 0.5469 - accuracy: 0.6712 - val_loss: 3.8943 - val_accuracy: 0.5493\n",
      "Epoch 12/50\n",
      "119997/119997 [==============================] - 4s 35us/sample - loss: 0.5449 - accuracy: 0.6748 - val_loss: 3.9162 - val_accuracy: 0.5590\n",
      "Epoch 13/50\n",
      "119997/119997 [==============================] - 4s 35us/sample - loss: 0.5424 - accuracy: 0.6750 - val_loss: 4.3897 - val_accuracy: 0.5636\n",
      "Epoch 14/50\n",
      "119997/119997 [==============================] - 5s 38us/sample - loss: 0.5420 - accuracy: 0.6738 - val_loss: 4.8737 - val_accuracy: 0.5673\n",
      "Epoch 15/50\n",
      "119997/119997 [==============================] - 4s 35us/sample - loss: 0.5434 - accuracy: 0.6730 - val_loss: 4.7869 - val_accuracy: 0.5658\n",
      "Epoch 16/50\n",
      "119997/119997 [==============================] - 4s 36us/sample - loss: 0.5379 - accuracy: 0.6785 - val_loss: 5.0018 - val_accuracy: 0.5535\n",
      "Epoch 17/50\n",
      "119997/119997 [==============================] - 4s 35us/sample - loss: 0.5392 - accuracy: 0.6763 - val_loss: 5.1975 - val_accuracy: 0.5556\n",
      "Epoch 18/50\n",
      "119997/119997 [==============================] - 4s 35us/sample - loss: 0.5391 - accuracy: 0.6777 - val_loss: 5.6493 - val_accuracy: 0.5713\n",
      "Epoch 19/50\n",
      "119997/119997 [==============================] - 4s 35us/sample - loss: 0.5381 - accuracy: 0.6778 - val_loss: 5.7637 - val_accuracy: 0.5645\n",
      "Epoch 20/50\n",
      "119997/119997 [==============================] - 4s 35us/sample - loss: 0.5384 - accuracy: 0.6769 - val_loss: 5.4835 - val_accuracy: 0.5556\n",
      "Epoch 21/50\n",
      "119997/119997 [==============================] - 4s 34us/sample - loss: 0.5353 - accuracy: 0.6796 - val_loss: 5.6108 - val_accuracy: 0.5924\n",
      "Epoch 22/50\n",
      "119997/119997 [==============================] - 4s 35us/sample - loss: 0.5351 - accuracy: 0.6797 - val_loss: 5.6998 - val_accuracy: 0.5358\n",
      "Epoch 23/50\n",
      "119997/119997 [==============================] - 4s 35us/sample - loss: 0.5353 - accuracy: 0.6812 - val_loss: 5.7712 - val_accuracy: 0.6096\n",
      "Epoch 24/50\n",
      "119997/119997 [==============================] - 4s 36us/sample - loss: 0.5306 - accuracy: 0.6842 - val_loss: 6.3539 - val_accuracy: 0.5721\n",
      "Epoch 25/50\n",
      "119997/119997 [==============================] - 4s 35us/sample - loss: 0.5301 - accuracy: 0.6839 - val_loss: 5.9502 - val_accuracy: 0.5974\n",
      "Epoch 26/50\n",
      "119997/119997 [==============================] - 4s 35us/sample - loss: 0.5333 - accuracy: 0.6820 - val_loss: 6.3699 - val_accuracy: 0.5616\n",
      "Epoch 27/50\n",
      "119997/119997 [==============================] - 4s 35us/sample - loss: 0.5351 - accuracy: 0.6802 - val_loss: 6.0663 - val_accuracy: 0.6136\n",
      "Epoch 28/50\n",
      "119997/119997 [==============================] - 4s 35us/sample - loss: 0.5370 - accuracy: 0.6782 - val_loss: 5.8872 - val_accuracy: 0.6157\n",
      "Epoch 29/50\n",
      "119997/119997 [==============================] - 4s 35us/sample - loss: 0.5344 - accuracy: 0.6806 - val_loss: 6.6549 - val_accuracy: 0.5688\n",
      "Epoch 30/50\n",
      "119997/119997 [==============================] - 4s 35us/sample - loss: 0.5350 - accuracy: 0.6801 - val_loss: 6.1776 - val_accuracy: 0.6215\n",
      "Epoch 31/50\n",
      "119997/119997 [==============================] - 4s 37us/sample - loss: 0.5295 - accuracy: 0.6839 - val_loss: 6.1071 - val_accuracy: 0.6059\n",
      "Epoch 32/50\n",
      "119997/119997 [==============================] - 4s 34us/sample - loss: 0.5257 - accuracy: 0.6882 - val_loss: 6.8203 - val_accuracy: 0.6262\n",
      "Epoch 33/50\n",
      "119997/119997 [==============================] - 4s 35us/sample - loss: 0.5198 - accuracy: 0.6935 - val_loss: 7.3853 - val_accuracy: 0.6139\n",
      "Epoch 34/50\n",
      "119997/119997 [==============================] - 4s 37us/sample - loss: 0.5152 - accuracy: 0.6972 - val_loss: 7.5246 - val_accuracy: 0.6231\n",
      "Epoch 35/50\n",
      "119997/119997 [==============================] - 4s 36us/sample - loss: 0.5151 - accuracy: 0.6974 - val_loss: 7.6765 - val_accuracy: 0.6298\n",
      "Epoch 36/50\n",
      "119997/119997 [==============================] - 4s 34us/sample - loss: 0.5130 - accuracy: 0.6989 - val_loss: 7.9228 - val_accuracy: 0.5640\n",
      "Epoch 37/50\n",
      "119997/119997 [==============================] - 4s 35us/sample - loss: 0.5098 - accuracy: 0.6997 - val_loss: 8.2758 - val_accuracy: 0.6361\n",
      "Epoch 38/50\n",
      "119997/119997 [==============================] - 4s 34us/sample - loss: 0.5136 - accuracy: 0.6976 - val_loss: 8.2352 - val_accuracy: 0.6314\n",
      "Epoch 39/50\n",
      "119997/119997 [==============================] - 4s 35us/sample - loss: 0.5108 - accuracy: 0.6989 - val_loss: 8.3109 - val_accuracy: 0.6149\n",
      "Epoch 40/50\n",
      "119997/119997 [==============================] - 4s 35us/sample - loss: 0.5079 - accuracy: 0.7031 - val_loss: 8.8812 - val_accuracy: 0.5654\n",
      "Epoch 41/50\n",
      "119997/119997 [==============================] - 4s 35us/sample - loss: 0.5088 - accuracy: 0.7021 - val_loss: 8.7868 - val_accuracy: 0.6085\n",
      "Epoch 42/50\n",
      "119997/119997 [==============================] - 4s 35us/sample - loss: 0.5072 - accuracy: 0.7027 - val_loss: 8.7013 - val_accuracy: 0.6360\n",
      "Epoch 43/50\n",
      "119997/119997 [==============================] - 4s 35us/sample - loss: 0.5021 - accuracy: 0.7066 - val_loss: 8.6205 - val_accuracy: 0.6097\n",
      "Epoch 44/50\n",
      "119997/119997 [==============================] - 4s 35us/sample - loss: 0.5066 - accuracy: 0.7024 - val_loss: 8.6715 - val_accuracy: 0.6240\n",
      "Epoch 45/50\n",
      "119997/119997 [==============================] - 4s 35us/sample - loss: 0.5062 - accuracy: 0.7029 - val_loss: 8.6698 - val_accuracy: 0.6419\n",
      "Epoch 46/50\n",
      "119997/119997 [==============================] - 4s 35us/sample - loss: 0.5072 - accuracy: 0.7020 - val_loss: 8.7215 - val_accuracy: 0.5955\n",
      "Epoch 47/50\n",
      "119997/119997 [==============================] - 4s 35us/sample - loss: 0.5120 - accuracy: 0.6993 - val_loss: 8.4573 - val_accuracy: 0.5855\n",
      "Epoch 48/50\n",
      "119997/119997 [==============================] - 4s 35us/sample - loss: 0.5052 - accuracy: 0.7037 - val_loss: 8.9826 - val_accuracy: 0.5803\n",
      "Epoch 49/50\n",
      "119997/119997 [==============================] - 4s 35us/sample - loss: 0.5063 - accuracy: 0.7019 - val_loss: 9.0544 - val_accuracy: 0.5719\n",
      "Epoch 50/50\n",
      "119997/119997 [==============================] - 4s 35us/sample - loss: 0.5068 - accuracy: 0.7019 - val_loss: 9.2286 - val_accuracy: 0.5709\n",
      "3999/3999 [==============================] - 0s 17us/sample - loss: 19.2694 - accuracy: 0.4929\n",
      "3999/3999 [==============================] - 0s 17us/sample - loss: 0.9635 - accuracy: 0.4981\n",
      "3999/3999 [==============================] - 0s 17us/sample - loss: 0.6545 - accuracy: 0.8450\n",
      "3-layers-32-nodes\n",
      "Train on 119997 samples, validate on 29997 samples\n",
      "Epoch 1/50\n",
      "119997/119997 [==============================] - 5s 39us/sample - loss: 0.6909 - accuracy: 0.5126 - val_loss: 0.6747 - val_accuracy: 0.5300\n",
      "Epoch 2/50\n",
      "119997/119997 [==============================] - 4s 37us/sample - loss: 0.6827 - accuracy: 0.5277 - val_loss: 0.7560 - val_accuracy: 0.5432\n",
      "Epoch 3/50\n",
      "119997/119997 [==============================] - 4s 35us/sample - loss: 0.6418 - accuracy: 0.5923 - val_loss: 1.1556 - val_accuracy: 0.5556\n",
      "Epoch 4/50\n",
      "119997/119997 [==============================] - 4s 35us/sample - loss: 0.5819 - accuracy: 0.6553 - val_loss: 1.9098 - val_accuracy: 0.5286\n",
      "Epoch 5/50\n",
      "119997/119997 [==============================] - 5s 41us/sample - loss: 0.5648 - accuracy: 0.6614 - val_loss: 2.6784 - val_accuracy: 0.5250\n",
      "Epoch 6/50\n",
      "119997/119997 [==============================] - 4s 36us/sample - loss: 0.5608 - accuracy: 0.6634 - val_loss: 2.1700 - val_accuracy: 0.5355\n",
      "Epoch 7/50\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "119997/119997 [==============================] - 4s 35us/sample - loss: 0.5572 - accuracy: 0.6654 - val_loss: 2.4257 - val_accuracy: 0.5396\n",
      "Epoch 8/50\n",
      "119997/119997 [==============================] - 4s 35us/sample - loss: 0.5543 - accuracy: 0.6653 - val_loss: 2.5890 - val_accuracy: 0.5586\n",
      "Epoch 9/50\n",
      "119997/119997 [==============================] - 4s 35us/sample - loss: 0.5533 - accuracy: 0.6663 - val_loss: 2.7997 - val_accuracy: 0.5332\n",
      "Epoch 10/50\n",
      "119997/119997 [==============================] - 4s 35us/sample - loss: 0.5499 - accuracy: 0.6698 - val_loss: 3.6632 - val_accuracy: 0.5543\n",
      "Epoch 11/50\n",
      "119997/119997 [==============================] - 4s 35us/sample - loss: 0.6091 - accuracy: 0.6170 - val_loss: 3.4861 - val_accuracy: 0.5350\n",
      "Epoch 12/50\n",
      "119997/119997 [==============================] - 4s 36us/sample - loss: 0.6022 - accuracy: 0.6232 - val_loss: 5.0744 - val_accuracy: 0.5459\n",
      "Epoch 13/50\n",
      "119997/119997 [==============================] - 4s 35us/sample - loss: 0.5937 - accuracy: 0.6341 - val_loss: 5.3019 - val_accuracy: 0.5819\n",
      "Epoch 14/50\n",
      "119997/119997 [==============================] - 4s 35us/sample - loss: 0.5971 - accuracy: 0.6261 - val_loss: 4.8634 - val_accuracy: 0.5691\n",
      "Epoch 15/50\n",
      "119997/119997 [==============================] - 4s 35us/sample - loss: 0.6092 - accuracy: 0.6080 - val_loss: 5.7104 - val_accuracy: 0.5445\n",
      "Epoch 16/50\n",
      "119997/119997 [==============================] - 4s 35us/sample - loss: 0.6123 - accuracy: 0.6046 - val_loss: 5.9405 - val_accuracy: 0.5812\n",
      "Epoch 17/50\n",
      "119997/119997 [==============================] - 4s 35us/sample - loss: 0.6004 - accuracy: 0.6164 - val_loss: 6.8114 - val_accuracy: 0.5544\n",
      "Epoch 18/50\n",
      "119997/119997 [==============================] - 4s 35us/sample - loss: 0.6041 - accuracy: 0.6136 - val_loss: 4.5158 - val_accuracy: 0.5734\n",
      "Epoch 19/50\n",
      "119997/119997 [==============================] - 4s 36us/sample - loss: 0.6061 - accuracy: 0.6098 - val_loss: 3.8321 - val_accuracy: 0.5383\n",
      "Epoch 20/50\n",
      "119997/119997 [==============================] - 4s 36us/sample - loss: 0.6098 - accuracy: 0.6034 - val_loss: 4.1001 - val_accuracy: 0.5629\n",
      "Epoch 21/50\n",
      "119997/119997 [==============================] - 4s 36us/sample - loss: 0.6339 - accuracy: 0.5722 - val_loss: 2.9216 - val_accuracy: 0.5555\n",
      "Epoch 22/50\n",
      "119997/119997 [==============================] - 4s 35us/sample - loss: 0.6377 - accuracy: 0.5683 - val_loss: 1.7227 - val_accuracy: 0.5569\n",
      "Epoch 23/50\n",
      "119997/119997 [==============================] - 4s 35us/sample - loss: 0.6469 - accuracy: 0.5561 - val_loss: 2.0343 - val_accuracy: 0.5425\n",
      "Epoch 24/50\n",
      "119997/119997 [==============================] - 4s 35us/sample - loss: 0.6532 - accuracy: 0.5461 - val_loss: 4.2519 - val_accuracy: 0.5564\n",
      "Epoch 25/50\n",
      "119997/119997 [==============================] - 4s 35us/sample - loss: 0.6407 - accuracy: 0.5631 - val_loss: 2.9244 - val_accuracy: 0.5389\n",
      "Epoch 26/50\n",
      "119997/119997 [==============================] - 4s 36us/sample - loss: 0.6334 - accuracy: 0.5760 - val_loss: 1.3860 - val_accuracy: 0.5248\n",
      "Epoch 27/50\n",
      "119997/119997 [==============================] - 4s 36us/sample - loss: 0.6157 - accuracy: 0.6110 - val_loss: 2.6167 - val_accuracy: 0.5788\n",
      "Epoch 28/50\n",
      "119997/119997 [==============================] - 4s 35us/sample - loss: 0.6220 - accuracy: 0.5859 - val_loss: 1.2723 - val_accuracy: 0.5985\n",
      "Epoch 29/50\n",
      "119997/119997 [==============================] - 4s 35us/sample - loss: 0.6084 - accuracy: 0.6037 - val_loss: 2.8777 - val_accuracy: 0.5831\n",
      "Epoch 30/50\n",
      "119997/119997 [==============================] - 4s 36us/sample - loss: 0.6065 - accuracy: 0.6058 - val_loss: 3.9573 - val_accuracy: 0.5578\n",
      "Epoch 31/50\n",
      "119997/119997 [==============================] - 4s 35us/sample - loss: 0.5763 - accuracy: 0.6432 - val_loss: 2.7236 - val_accuracy: 0.5778\n",
      "Epoch 32/50\n",
      "119997/119997 [==============================] - 4s 35us/sample - loss: 0.5568 - accuracy: 0.6675 - val_loss: 2.1401 - val_accuracy: 0.5683\n",
      "Epoch 33/50\n",
      "119997/119997 [==============================] - 4s 36us/sample - loss: 0.5415 - accuracy: 0.6803 - val_loss: 3.0553 - val_accuracy: 0.5687\n",
      "Epoch 34/50\n",
      "119997/119997 [==============================] - 4s 36us/sample - loss: 0.5725 - accuracy: 0.6419 - val_loss: 2.7042 - val_accuracy: 0.5117\n",
      "Epoch 35/50\n",
      "119997/119997 [==============================] - 4s 35us/sample - loss: 0.5544 - accuracy: 0.6644 - val_loss: 3.3440 - val_accuracy: 0.5871\n",
      "Epoch 36/50\n",
      "119997/119997 [==============================] - 4s 35us/sample - loss: 0.5462 - accuracy: 0.6690 - val_loss: 3.6657 - val_accuracy: 0.5987\n",
      "Epoch 37/50\n",
      "119997/119997 [==============================] - 4s 35us/sample - loss: 0.5560 - accuracy: 0.6623 - val_loss: 4.1329 - val_accuracy: 0.5790\n",
      "Epoch 38/50\n",
      "119997/119997 [==============================] - 4s 35us/sample - loss: 0.5331 - accuracy: 0.6866 - val_loss: 3.8791 - val_accuracy: 0.5743\n",
      "Epoch 39/50\n",
      "119997/119997 [==============================] - 4s 35us/sample - loss: 0.5274 - accuracy: 0.6943 - val_loss: 4.2271 - val_accuracy: 0.6084\n",
      "Epoch 40/50\n",
      "119997/119997 [==============================] - 4s 36us/sample - loss: 0.5466 - accuracy: 0.6692 - val_loss: 3.7618 - val_accuracy: 0.5812\n",
      "Epoch 41/50\n",
      "119997/119997 [==============================] - 4s 36us/sample - loss: 0.5232 - accuracy: 0.6950 - val_loss: 3.8601 - val_accuracy: 0.5941\n",
      "Epoch 42/50\n",
      "119997/119997 [==============================] - 4s 35us/sample - loss: 0.5161 - accuracy: 0.6993 - val_loss: 4.2998 - val_accuracy: 0.5727\n",
      "Epoch 43/50\n",
      "119997/119997 [==============================] - 4s 35us/sample - loss: 0.5383 - accuracy: 0.6773 - val_loss: 4.2521 - val_accuracy: 0.5977\n",
      "Epoch 44/50\n",
      "119997/119997 [==============================] - 4s 35us/sample - loss: 0.5146 - accuracy: 0.7001 - val_loss: 4.1885 - val_accuracy: 0.5903\n",
      "Epoch 45/50\n",
      "119997/119997 [==============================] - 4s 35us/sample - loss: 0.5317 - accuracy: 0.6826 - val_loss: 3.9967 - val_accuracy: 0.5722\n",
      "Epoch 46/50\n",
      "119997/119997 [==============================] - 4s 35us/sample - loss: 0.5436 - accuracy: 0.6787 - val_loss: 3.5941 - val_accuracy: 0.5439\n",
      "Epoch 47/50\n",
      "119997/119997 [==============================] - 4s 36us/sample - loss: 0.5096 - accuracy: 0.7066 - val_loss: 2.5397 - val_accuracy: 0.5399\n",
      "Epoch 48/50\n",
      "119997/119997 [==============================] - 4s 36us/sample - loss: 0.5102 - accuracy: 0.7049 - val_loss: 5.7803 - val_accuracy: 0.5574\n",
      "Epoch 49/50\n",
      "119997/119997 [==============================] - 4s 35us/sample - loss: 0.5196 - accuracy: 0.6957 - val_loss: 4.5172 - val_accuracy: 0.5755\n",
      "Epoch 50/50\n",
      "119997/119997 [==============================] - 4s 35us/sample - loss: 0.5105 - accuracy: 0.7041 - val_loss: 3.2619 - val_accuracy: 0.5407\n",
      "3999/3999 [==============================] - 0s 17us/sample - loss: 26.1212 - accuracy: 0.4456\n",
      "3999/3999 [==============================] - 0s 21us/sample - loss: 1.2755 - accuracy: 0.5001\n",
      "3999/3999 [==============================] - 0s 17us/sample - loss: 0.8597 - accuracy: 0.6004\n",
      "3-layers-64-nodes\n",
      "Train on 119997 samples, validate on 29997 samples\n",
      "Epoch 1/50\n",
      "119997/119997 [==============================] - 5s 40us/sample - loss: 0.6906 - accuracy: 0.5102 - val_loss: 0.6811 - val_accuracy: 0.4952\n",
      "Epoch 2/50\n",
      "119997/119997 [==============================] - 4s 35us/sample - loss: 0.6872 - accuracy: 0.5180 - val_loss: 0.6976 - val_accuracy: 0.5354\n",
      "Epoch 3/50\n",
      "119997/119997 [==============================] - 4s 36us/sample - loss: 0.6657 - accuracy: 0.5637 - val_loss: 0.9575 - val_accuracy: 0.4665\n",
      "Epoch 4/50\n",
      "119997/119997 [==============================] - 4s 37us/sample - loss: 0.5902 - accuracy: 0.6400 - val_loss: 1.2148 - val_accuracy: 0.4828\n",
      "Epoch 5/50\n",
      "119997/119997 [==============================] - 4s 36us/sample - loss: 0.5753 - accuracy: 0.6519 - val_loss: 1.3773 - val_accuracy: 0.4948\n",
      "Epoch 6/50\n",
      "119997/119997 [==============================] - 4s 35us/sample - loss: 0.5712 - accuracy: 0.6524 - val_loss: 1.5693 - val_accuracy: 0.4937\n",
      "Epoch 7/50\n",
      "119997/119997 [==============================] - 4s 36us/sample - loss: 0.5744 - accuracy: 0.6493 - val_loss: 1.3296 - val_accuracy: 0.4955\n",
      "Epoch 8/50\n",
      "119997/119997 [==============================] - 5s 38us/sample - loss: 0.5700 - accuracy: 0.6541 - val_loss: 1.7674 - val_accuracy: 0.4952\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 9/50\n",
      "119997/119997 [==============================] - 4s 36us/sample - loss: 0.5803 - accuracy: 0.6435 - val_loss: 1.5980 - val_accuracy: 0.4985\n",
      "Epoch 10/50\n",
      "119997/119997 [==============================] - 4s 35us/sample - loss: 0.6005 - accuracy: 0.6215 - val_loss: 1.5110 - val_accuracy: 0.4993\n",
      "Epoch 11/50\n",
      "119997/119997 [==============================] - 4s 37us/sample - loss: 0.5815 - accuracy: 0.6421 - val_loss: 1.8409 - val_accuracy: 0.5053\n",
      "Epoch 12/50\n",
      "119997/119997 [==============================] - 4s 36us/sample - loss: 0.5668 - accuracy: 0.6566 - val_loss: 1.6294 - val_accuracy: 0.5052\n",
      "Epoch 13/50\n",
      "119997/119997 [==============================] - 4s 36us/sample - loss: 0.5680 - accuracy: 0.6561 - val_loss: 1.6803 - val_accuracy: 0.5043\n",
      "Epoch 14/50\n",
      "119997/119997 [==============================] - 4s 35us/sample - loss: 0.5556 - accuracy: 0.6666 - val_loss: 1.5418 - val_accuracy: 0.5186\n",
      "Epoch 15/50\n",
      "119997/119997 [==============================] - 4s 35us/sample - loss: 0.5571 - accuracy: 0.6648 - val_loss: 1.5507 - val_accuracy: 0.5097\n",
      "Epoch 16/50\n",
      "119997/119997 [==============================] - 4s 35us/sample - loss: 0.5536 - accuracy: 0.6713 - val_loss: 1.5906 - val_accuracy: 0.5289\n",
      "Epoch 17/50\n",
      "119997/119997 [==============================] - 4s 35us/sample - loss: 0.5504 - accuracy: 0.6768 - val_loss: 1.1577 - val_accuracy: 0.5114\n",
      "Epoch 18/50\n",
      "119997/119997 [==============================] - 4s 37us/sample - loss: 0.5451 - accuracy: 0.6885 - val_loss: 1.2215 - val_accuracy: 0.5212\n",
      "Epoch 19/50\n",
      "119997/119997 [==============================] - 4s 36us/sample - loss: 0.5317 - accuracy: 0.7053 - val_loss: 1.2922 - val_accuracy: 0.5040\n",
      "Epoch 20/50\n",
      "119997/119997 [==============================] - 4s 37us/sample - loss: 0.5253 - accuracy: 0.7164 - val_loss: 1.3409 - val_accuracy: 0.5278\n",
      "Epoch 21/50\n",
      "119997/119997 [==============================] - 4s 37us/sample - loss: 0.5006 - accuracy: 0.7399 - val_loss: 1.3018 - val_accuracy: 0.5338\n",
      "Epoch 22/50\n",
      "119997/119997 [==============================] - 5s 40us/sample - loss: 0.4929 - accuracy: 0.7440 - val_loss: 1.3675 - val_accuracy: 0.5410\n",
      "Epoch 23/50\n",
      "119997/119997 [==============================] - 4s 37us/sample - loss: 0.5071 - accuracy: 0.7252 - val_loss: 1.2686 - val_accuracy: 0.5582\n",
      "Epoch 24/50\n",
      "119997/119997 [==============================] - 4s 35us/sample - loss: 0.4766 - accuracy: 0.7521 - val_loss: 1.5022 - val_accuracy: 0.5364\n",
      "Epoch 25/50\n",
      "119997/119997 [==============================] - 4s 36us/sample - loss: 0.4749 - accuracy: 0.7523 - val_loss: 1.3336 - val_accuracy: 0.5528\n",
      "Epoch 26/50\n",
      "119997/119997 [==============================] - 4s 36us/sample - loss: 0.4636 - accuracy: 0.7612 - val_loss: 1.0564 - val_accuracy: 0.5406\n",
      "Epoch 27/50\n",
      "119997/119997 [==============================] - 4s 36us/sample - loss: 0.4856 - accuracy: 0.7375 - val_loss: 1.2207 - val_accuracy: 0.5590\n",
      "Epoch 28/50\n",
      "119997/119997 [==============================] - 4s 35us/sample - loss: 0.4729 - accuracy: 0.7508 - val_loss: 1.2932 - val_accuracy: 0.5273\n",
      "Epoch 29/50\n",
      "119997/119997 [==============================] - 4s 36us/sample - loss: 0.4706 - accuracy: 0.7530 - val_loss: 1.2937 - val_accuracy: 0.5477\n",
      "Epoch 30/50\n",
      "119997/119997 [==============================] - 4s 36us/sample - loss: 0.4839 - accuracy: 0.7363 - val_loss: 1.3541 - val_accuracy: 0.5460\n",
      "Epoch 31/50\n",
      "119997/119997 [==============================] - 4s 36us/sample - loss: 0.4771 - accuracy: 0.7410 - val_loss: 1.0531 - val_accuracy: 0.5046\n",
      "Epoch 32/50\n",
      "119997/119997 [==============================] - 4s 36us/sample - loss: 0.4671 - accuracy: 0.7532 - val_loss: 1.0874 - val_accuracy: 0.5699\n",
      "Epoch 33/50\n",
      "119997/119997 [==============================] - 4s 36us/sample - loss: 0.4691 - accuracy: 0.7519 - val_loss: 1.1770 - val_accuracy: 0.5521\n",
      "Epoch 34/50\n",
      "119997/119997 [==============================] - 4s 35us/sample - loss: 0.4719 - accuracy: 0.7454 - val_loss: 1.2248 - val_accuracy: 0.5492\n",
      "Epoch 35/50\n",
      "119997/119997 [==============================] - 4s 37us/sample - loss: 0.4475 - accuracy: 0.7687 - val_loss: 0.9936 - val_accuracy: 0.5656\n",
      "Epoch 36/50\n",
      "119997/119997 [==============================] - 4s 36us/sample - loss: 0.4565 - accuracy: 0.7600 - val_loss: 1.0793 - val_accuracy: 0.5837\n",
      "Epoch 37/50\n",
      "119997/119997 [==============================] - 4s 35us/sample - loss: 0.4573 - accuracy: 0.7604 - val_loss: 1.1496 - val_accuracy: 0.5780\n",
      "Epoch 38/50\n",
      "119997/119997 [==============================] - 4s 35us/sample - loss: 0.4807 - accuracy: 0.7346 - val_loss: 1.0571 - val_accuracy: 0.5691\n",
      "Epoch 39/50\n",
      "119997/119997 [==============================] - 4s 36us/sample - loss: 0.4564 - accuracy: 0.7607 - val_loss: 1.2695 - val_accuracy: 0.5637\n",
      "Epoch 40/50\n",
      "119997/119997 [==============================] - 4s 36us/sample - loss: 0.4624 - accuracy: 0.7538 - val_loss: 1.2371 - val_accuracy: 0.5603\n",
      "Epoch 41/50\n",
      "119997/119997 [==============================] - 5s 39us/sample - loss: 0.4664 - accuracy: 0.7482 - val_loss: 1.0262 - val_accuracy: 0.5491\n",
      "Epoch 42/50\n",
      "119997/119997 [==============================] - 4s 36us/sample - loss: 0.4642 - accuracy: 0.7511 - val_loss: 1.0483 - val_accuracy: 0.5642\n",
      "Epoch 43/50\n",
      "119997/119997 [==============================] - 4s 35us/sample - loss: 0.4458 - accuracy: 0.7732 - val_loss: 1.0091 - val_accuracy: 0.5526\n",
      "Epoch 44/50\n",
      "119997/119997 [==============================] - 4s 35us/sample - loss: 0.5064 - accuracy: 0.7091 - val_loss: 1.1258 - val_accuracy: 0.5203\n",
      "Epoch 45/50\n",
      "119997/119997 [==============================] - 4s 35us/sample - loss: 0.4902 - accuracy: 0.7257 - val_loss: 0.9176 - val_accuracy: 0.5584\n",
      "Epoch 46/50\n",
      "119997/119997 [==============================] - 4s 36us/sample - loss: 0.4518 - accuracy: 0.7629 - val_loss: 1.2100 - val_accuracy: 0.5737\n",
      "Epoch 47/50\n",
      "119997/119997 [==============================] - 4s 36us/sample - loss: 0.4712 - accuracy: 0.7434 - val_loss: 1.2751 - val_accuracy: 0.5578\n",
      "Epoch 48/50\n",
      "119997/119997 [==============================] - 5s 38us/sample - loss: 0.4629 - accuracy: 0.7532 - val_loss: 1.3113 - val_accuracy: 0.5753\n",
      "Epoch 49/50\n",
      "119997/119997 [==============================] - 5s 45us/sample - loss: 0.4545 - accuracy: 0.7651 - val_loss: 1.2489 - val_accuracy: 0.5593\n",
      "Epoch 50/50\n",
      "119997/119997 [==============================] - 6s 47us/sample - loss: 0.4405 - accuracy: 0.7780 - val_loss: 1.3901 - val_accuracy: 0.5180\n",
      "3999/3999 [==============================] - 0s 15us/sample - loss: 6.8113 - accuracy: 0.4066\n",
      "3999/3999 [==============================] - 0s 15us/sample - loss: 0.6035 - accuracy: 0.5131\n",
      "3999/3999 [==============================] - 0s 15us/sample - loss: 3.7259 - accuracy: 0.6429\n",
      "3-layers-128-nodes\n",
      "Train on 119997 samples, validate on 29997 samples\n",
      "Epoch 1/50\n",
      "119997/119997 [==============================] - 5s 45us/sample - loss: 0.6901 - accuracy: 0.5098 - val_loss: 0.6585 - val_accuracy: 0.5666\n",
      "Epoch 2/50\n",
      "119997/119997 [==============================] - 5s 40us/sample - loss: 0.6805 - accuracy: 0.5310 - val_loss: 0.9532 - val_accuracy: 0.5209\n",
      "Epoch 3/50\n",
      "119997/119997 [==============================] - 5s 39us/sample - loss: 0.6246 - accuracy: 0.6109 - val_loss: 1.9306 - val_accuracy: 0.5206\n",
      "Epoch 4/50\n",
      "119997/119997 [==============================] - 5s 39us/sample - loss: 0.5562 - accuracy: 0.6799 - val_loss: 3.6278 - val_accuracy: 0.5134\n",
      "Epoch 5/50\n",
      "119997/119997 [==============================] - 5s 39us/sample - loss: 0.5166 - accuracy: 0.7191 - val_loss: 7.1385 - val_accuracy: 0.4976\n",
      "Epoch 6/50\n",
      "119997/119997 [==============================] - 5s 39us/sample - loss: 0.4999 - accuracy: 0.7346 - val_loss: 9.5127 - val_accuracy: 0.4931\n",
      "Epoch 7/50\n",
      "119997/119997 [==============================] - 5s 39us/sample - loss: 0.4902 - accuracy: 0.7434 - val_loss: 10.4669 - val_accuracy: 0.4903\n",
      "Epoch 8/50\n",
      "119997/119997 [==============================] - 5s 39us/sample - loss: 0.4861 - accuracy: 0.7457 - val_loss: 9.8790 - val_accuracy: 0.4990\n",
      "Epoch 9/50\n",
      "119997/119997 [==============================] - 5s 39us/sample - loss: 0.4772 - accuracy: 0.7512 - val_loss: 11.3220 - val_accuracy: 0.4887\n",
      "Epoch 10/50\n",
      "119997/119997 [==============================] - 5s 38us/sample - loss: 0.4695 - accuracy: 0.7569 - val_loss: 11.6306 - val_accuracy: 0.4885\n",
      "Epoch 11/50\n",
      "119997/119997 [==============================] - 5s 40us/sample - loss: 0.4670 - accuracy: 0.7588 - val_loss: 11.8942 - val_accuracy: 0.4865\n",
      "Epoch 12/50\n",
      "119997/119997 [==============================] - 5s 39us/sample - loss: 0.4620 - accuracy: 0.7615 - val_loss: 11.4786 - val_accuracy: 0.4920\n",
      "Epoch 13/50\n",
      "119997/119997 [==============================] - 5s 39us/sample - loss: 0.4610 - accuracy: 0.7615 - val_loss: 13.1222 - val_accuracy: 0.4938\n",
      "Epoch 14/50\n",
      "119997/119997 [==============================] - 5s 41us/sample - loss: 0.4497 - accuracy: 0.7720 - val_loss: 12.9388 - val_accuracy: 0.4855\n",
      "Epoch 15/50\n",
      "119997/119997 [==============================] - 5s 39us/sample - loss: 0.4506 - accuracy: 0.7680 - val_loss: 13.7212 - val_accuracy: 0.5020\n",
      "Epoch 16/50\n",
      "119997/119997 [==============================] - 5s 39us/sample - loss: 0.4495 - accuracy: 0.7695 - val_loss: 14.7687 - val_accuracy: 0.4959\n",
      "Epoch 17/50\n",
      "119997/119997 [==============================] - 5s 39us/sample - loss: 0.4509 - accuracy: 0.7678 - val_loss: 13.9458 - val_accuracy: 0.4950\n",
      "Epoch 18/50\n",
      "119997/119997 [==============================] - 5s 40us/sample - loss: 0.4424 - accuracy: 0.7737 - val_loss: 12.5111 - val_accuracy: 0.5039\n",
      "Epoch 19/50\n",
      "119997/119997 [==============================] - 5s 40us/sample - loss: 0.4432 - accuracy: 0.7738 - val_loss: 13.9094 - val_accuracy: 0.4995\n",
      "Epoch 20/50\n",
      "119997/119997 [==============================] - 5s 40us/sample - loss: 0.4391 - accuracy: 0.7755 - val_loss: 14.0853 - val_accuracy: 0.4988\n",
      "Epoch 21/50\n",
      "119997/119997 [==============================] - 5s 40us/sample - loss: 0.4413 - accuracy: 0.7739 - val_loss: 14.4939 - val_accuracy: 0.4968\n",
      "Epoch 22/50\n",
      "119997/119997 [==============================] - 5s 40us/sample - loss: 0.4499 - accuracy: 0.7679 - val_loss: 14.2117 - val_accuracy: 0.5007\n",
      "Epoch 23/50\n",
      "119997/119997 [==============================] - 5s 40us/sample - loss: 0.4400 - accuracy: 0.7752 - val_loss: 14.9260 - val_accuracy: 0.5007\n",
      "Epoch 24/50\n",
      "119997/119997 [==============================] - 5s 40us/sample - loss: 0.4268 - accuracy: 0.7836 - val_loss: 14.5659 - val_accuracy: 0.5009\n",
      "Epoch 25/50\n",
      "119997/119997 [==============================] - 5s 40us/sample - loss: 0.4401 - accuracy: 0.7750 - val_loss: 14.8975 - val_accuracy: 0.5020\n",
      "Epoch 26/50\n",
      "119997/119997 [==============================] - 5s 40us/sample - loss: 0.4321 - accuracy: 0.7801 - val_loss: 14.4530 - val_accuracy: 0.4982\n",
      "Epoch 27/50\n",
      "119997/119997 [==============================] - 5s 40us/sample - loss: 0.4258 - accuracy: 0.7851 - val_loss: 15.1920 - val_accuracy: 0.4968\n",
      "Epoch 28/50\n",
      "119997/119997 [==============================] - 5s 40us/sample - loss: 0.4284 - accuracy: 0.7831 - val_loss: 15.8331 - val_accuracy: 0.4932\n",
      "Epoch 29/50\n",
      "119997/119997 [==============================] - 5s 40us/sample - loss: 0.4346 - accuracy: 0.7798 - val_loss: 15.1026 - val_accuracy: 0.4923\n",
      "Epoch 30/50\n",
      "119997/119997 [==============================] - 5s 40us/sample - loss: 0.4269 - accuracy: 0.7841 - val_loss: 15.6308 - val_accuracy: 0.4976\n",
      "Epoch 31/50\n",
      "119997/119997 [==============================] - 5s 40us/sample - loss: 0.4420 - accuracy: 0.7730 - val_loss: 15.3266 - val_accuracy: 0.4936\n",
      "Epoch 32/50\n",
      "119997/119997 [==============================] - 5s 40us/sample - loss: 0.4287 - accuracy: 0.7825 - val_loss: 17.0434 - val_accuracy: 0.4976\n",
      "Epoch 33/50\n",
      "119997/119997 [==============================] - 5s 41us/sample - loss: 0.4263 - accuracy: 0.7853 - val_loss: 14.6290 - val_accuracy: 0.5024\n",
      "Epoch 34/50\n",
      "119997/119997 [==============================] - 5s 40us/sample - loss: 0.4232 - accuracy: 0.7870 - val_loss: 15.1291 - val_accuracy: 0.5005\n",
      "Epoch 35/50\n",
      "119997/119997 [==============================] - 5s 40us/sample - loss: 0.4343 - accuracy: 0.7807 - val_loss: 16.1687 - val_accuracy: 0.4971\n",
      "Epoch 36/50\n",
      "119997/119997 [==============================] - 5s 41us/sample - loss: 0.4303 - accuracy: 0.7833 - val_loss: 14.9668 - val_accuracy: 0.4974\n",
      "Epoch 37/50\n",
      "119997/119997 [==============================] - 5s 41us/sample - loss: 0.4276 - accuracy: 0.7847 - val_loss: 15.8915 - val_accuracy: 0.4994\n",
      "Epoch 38/50\n",
      "119997/119997 [==============================] - 5s 39us/sample - loss: 0.4286 - accuracy: 0.7840 - val_loss: 15.4653 - val_accuracy: 0.4982\n",
      "Epoch 39/50\n",
      "119997/119997 [==============================] - 5s 41us/sample - loss: 0.4256 - accuracy: 0.7853 - val_loss: 13.5187 - val_accuracy: 0.5042\n",
      "Epoch 40/50\n",
      "119997/119997 [==============================] - 5s 39us/sample - loss: 0.4192 - accuracy: 0.7917 - val_loss: 15.6056 - val_accuracy: 0.4986\n",
      "Epoch 41/50\n",
      "119997/119997 [==============================] - 5s 39us/sample - loss: 0.4233 - accuracy: 0.7882 - val_loss: 16.9564 - val_accuracy: 0.5075\n",
      "Epoch 42/50\n",
      "119997/119997 [==============================] - 5s 39us/sample - loss: 0.4289 - accuracy: 0.7825 - val_loss: 15.4285 - val_accuracy: 0.5044\n",
      "Epoch 43/50\n",
      "119997/119997 [==============================] - 5s 40us/sample - loss: 0.4258 - accuracy: 0.7879 - val_loss: 14.6255 - val_accuracy: 0.5067\n",
      "Epoch 44/50\n",
      "119997/119997 [==============================] - 5s 39us/sample - loss: 0.4187 - accuracy: 0.7910 - val_loss: 15.0590 - val_accuracy: 0.5028\n",
      "Epoch 45/50\n",
      "119997/119997 [==============================] - 5s 40us/sample - loss: 0.4236 - accuracy: 0.7895 - val_loss: 15.9943 - val_accuracy: 0.5043\n",
      "Epoch 46/50\n",
      "119997/119997 [==============================] - 5s 40us/sample - loss: 0.4204 - accuracy: 0.7896 - val_loss: 15.8257 - val_accuracy: 0.5042\n",
      "Epoch 47/50\n",
      "119997/119997 [==============================] - 5s 39us/sample - loss: 0.4366 - accuracy: 0.7770 - val_loss: 14.0621 - val_accuracy: 0.4994\n",
      "Epoch 48/50\n",
      "119997/119997 [==============================] - 5s 40us/sample - loss: 0.4272 - accuracy: 0.7842 - val_loss: 15.2620 - val_accuracy: 0.5060\n",
      "Epoch 49/50\n",
      "119997/119997 [==============================] - 5s 39us/sample - loss: 0.4190 - accuracy: 0.7921 - val_loss: 14.7345 - val_accuracy: 0.5055\n",
      "Epoch 50/50\n",
      "119997/119997 [==============================] - 5s 40us/sample - loss: 0.4113 - accuracy: 0.7965 - val_loss: 15.5300 - val_accuracy: 0.5048\n",
      "3999/3999 [==============================] - 0s 19us/sample - loss: 5.3908 - accuracy: 0.4669\n",
      "3999/3999 [==============================] - 0s 19us/sample - loss: 19.7196 - accuracy: 0.5629\n",
      "3999/3999 [==============================] - 0s 19us/sample - loss: 23.9797 - accuracy: 0.6229\n",
      "4-layers-16-nodes\n",
      "Train on 119997 samples, validate on 29997 samples\n",
      "Epoch 1/50\n",
      "119997/119997 [==============================] - 5s 41us/sample - loss: 0.6909 - accuracy: 0.5051 - val_loss: 0.6559 - val_accuracy: 0.5248\n",
      "Epoch 2/50\n",
      "119997/119997 [==============================] - 5s 38us/sample - loss: 0.6894 - accuracy: 0.5033 - val_loss: 0.6575 - val_accuracy: 0.5630\n",
      "Epoch 3/50\n",
      "119997/119997 [==============================] - 4s 37us/sample - loss: 0.6892 - accuracy: 0.5036 - val_loss: 0.6475 - val_accuracy: 0.5925\n",
      "Epoch 4/50\n",
      "119997/119997 [==============================] - 4s 37us/sample - loss: 0.6887 - accuracy: 0.5032 - val_loss: 0.6621 - val_accuracy: 0.5687\n",
      "Epoch 5/50\n",
      "119997/119997 [==============================] - 5s 38us/sample - loss: 0.6868 - accuracy: 0.5128 - val_loss: 0.6859 - val_accuracy: 0.5586\n",
      "Epoch 6/50\n",
      "119997/119997 [==============================] - 5s 38us/sample - loss: 0.6822 - accuracy: 0.5267 - val_loss: 0.7406 - val_accuracy: 0.5860\n",
      "Epoch 7/50\n",
      "119997/119997 [==============================] - 4s 37us/sample - loss: 0.6796 - accuracy: 0.5299 - val_loss: 0.8974 - val_accuracy: 0.5388\n",
      "Epoch 8/50\n",
      "119997/119997 [==============================] - 4s 37us/sample - loss: 0.6751 - accuracy: 0.5349 - val_loss: 0.9371 - val_accuracy: 0.4959\n",
      "Epoch 9/50\n",
      "119997/119997 [==============================] - 5s 39us/sample - loss: 0.6686 - accuracy: 0.5417 - val_loss: 1.2028 - val_accuracy: 0.5079\n",
      "Epoch 10/50\n",
      "119997/119997 [==============================] - 4s 37us/sample - loss: 0.6635 - accuracy: 0.5474 - val_loss: 1.4373 - val_accuracy: 0.5041\n",
      "Epoch 11/50\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "119997/119997 [==============================] - 4s 37us/sample - loss: 0.6450 - accuracy: 0.5716 - val_loss: 1.8264 - val_accuracy: 0.5111\n",
      "Epoch 12/50\n",
      "119997/119997 [==============================] - 5s 39us/sample - loss: 0.6543 - accuracy: 0.5541 - val_loss: 1.8495 - val_accuracy: 0.5181\n",
      "Epoch 13/50\n",
      "119997/119997 [==============================] - 4s 37us/sample - loss: 0.6451 - accuracy: 0.5621 - val_loss: 1.9545 - val_accuracy: 0.5865\n",
      "Epoch 14/50\n",
      "119997/119997 [==============================] - 4s 37us/sample - loss: 0.6508 - accuracy: 0.5592 - val_loss: 2.0476 - val_accuracy: 0.5217\n",
      "Epoch 15/50\n",
      "119997/119997 [==============================] - 5s 38us/sample - loss: 0.6314 - accuracy: 0.5895 - val_loss: 2.1293 - val_accuracy: 0.5215\n",
      "Epoch 16/50\n",
      "119997/119997 [==============================] - 4s 37us/sample - loss: 0.6266 - accuracy: 0.5933 - val_loss: 1.5930 - val_accuracy: 0.4912\n",
      "Epoch 17/50\n",
      "119997/119997 [==============================] - 4s 37us/sample - loss: 0.6613 - accuracy: 0.5440 - val_loss: 2.1916 - val_accuracy: 0.4972\n",
      "Epoch 18/50\n",
      "119997/119997 [==============================] - 4s 37us/sample - loss: 0.6783 - accuracy: 0.5194 - val_loss: 0.9081 - val_accuracy: 0.5880\n",
      "Epoch 19/50\n",
      "119997/119997 [==============================] - 4s 37us/sample - loss: 0.6670 - accuracy: 0.5344 - val_loss: 0.8612 - val_accuracy: 0.5801\n",
      "Epoch 20/50\n",
      "119997/119997 [==============================] - 4s 37us/sample - loss: 0.6470 - accuracy: 0.5615 - val_loss: 1.3305 - val_accuracy: 0.5399\n",
      "Epoch 21/50\n",
      "119997/119997 [==============================] - 4s 37us/sample - loss: 0.6042 - accuracy: 0.6134 - val_loss: 2.6493 - val_accuracy: 0.5773\n",
      "Epoch 22/50\n",
      "119997/119997 [==============================] - 5s 38us/sample - loss: 0.5621 - accuracy: 0.6623 - val_loss: 3.6364 - val_accuracy: 0.5733\n",
      "Epoch 23/50\n",
      "119997/119997 [==============================] - 5s 38us/sample - loss: 0.5655 - accuracy: 0.6543 - val_loss: 2.3167 - val_accuracy: 0.4880\n",
      "Epoch 24/50\n",
      "119997/119997 [==============================] - 4s 37us/sample - loss: 0.5807 - accuracy: 0.6385 - val_loss: 2.4267 - val_accuracy: 0.4887\n",
      "Epoch 25/50\n",
      "119997/119997 [==============================] - 4s 37us/sample - loss: 0.5544 - accuracy: 0.6631 - val_loss: 3.2237 - val_accuracy: 0.4942\n",
      "Epoch 26/50\n",
      "119997/119997 [==============================] - 4s 37us/sample - loss: 0.5313 - accuracy: 0.6836 - val_loss: 4.9893 - val_accuracy: 0.5254\n",
      "Epoch 27/50\n",
      "119997/119997 [==============================] - 4s 37us/sample - loss: 0.5366 - accuracy: 0.6759 - val_loss: 3.5737 - val_accuracy: 0.5421\n",
      "Epoch 28/50\n",
      "119997/119997 [==============================] - 4s 37us/sample - loss: 0.5234 - accuracy: 0.6872 - val_loss: 5.3031 - val_accuracy: 0.5228\n",
      "Epoch 29/50\n",
      "119997/119997 [==============================] - 4s 37us/sample - loss: 0.5286 - accuracy: 0.6836 - val_loss: 2.2073 - val_accuracy: 0.4886\n",
      "Epoch 30/50\n",
      "119997/119997 [==============================] - 4s 37us/sample - loss: 0.5522 - accuracy: 0.6534 - val_loss: 5.3585 - val_accuracy: 0.5287\n",
      "Epoch 31/50\n",
      "119997/119997 [==============================] - 4s 37us/sample - loss: 0.5184 - accuracy: 0.6895 - val_loss: 4.3624 - val_accuracy: 0.5255\n",
      "Epoch 32/50\n",
      "119997/119997 [==============================] - 4s 37us/sample - loss: 0.5200 - accuracy: 0.6913 - val_loss: 3.2753 - val_accuracy: 0.4941\n",
      "Epoch 33/50\n",
      "119997/119997 [==============================] - 4s 37us/sample - loss: 0.5285 - accuracy: 0.6846 - val_loss: 2.1812 - val_accuracy: 0.5358\n",
      "Epoch 34/50\n",
      "119997/119997 [==============================] - 4s 37us/sample - loss: 0.5552 - accuracy: 0.6498 - val_loss: 4.1841 - val_accuracy: 0.5723\n",
      "Epoch 35/50\n",
      "119997/119997 [==============================] - 5s 38us/sample - loss: 0.5371 - accuracy: 0.6731 - val_loss: 6.0916 - val_accuracy: 0.5285\n",
      "Epoch 36/50\n",
      "119997/119997 [==============================] - 4s 37us/sample - loss: 0.5268 - accuracy: 0.6832 - val_loss: 3.7618 - val_accuracy: 0.5034\n",
      "Epoch 37/50\n",
      "119997/119997 [==============================] - 4s 37us/sample - loss: 0.5367 - accuracy: 0.6724 - val_loss: 5.9773 - val_accuracy: 0.5290\n",
      "Epoch 38/50\n",
      "119997/119997 [==============================] - 4s 37us/sample - loss: 0.5319 - accuracy: 0.6779 - val_loss: 4.5901 - val_accuracy: 0.5359\n",
      "Epoch 39/50\n",
      "119997/119997 [==============================] - 5s 39us/sample - loss: 0.5482 - accuracy: 0.6553 - val_loss: 4.4359 - val_accuracy: 0.5265\n",
      "Epoch 40/50\n",
      "119997/119997 [==============================] - 4s 37us/sample - loss: 0.5280 - accuracy: 0.6826 - val_loss: 4.5532 - val_accuracy: 0.5292\n",
      "Epoch 41/50\n",
      "119997/119997 [==============================] - 4s 37us/sample - loss: 0.5180 - accuracy: 0.6933 - val_loss: 5.7584 - val_accuracy: 0.5289\n",
      "Epoch 42/50\n",
      "119997/119997 [==============================] - 4s 37us/sample - loss: 0.5417 - accuracy: 0.6738 - val_loss: 6.5049 - val_accuracy: 0.5555\n",
      "Epoch 43/50\n",
      "119997/119997 [==============================] - 4s 37us/sample - loss: 0.5400 - accuracy: 0.6734 - val_loss: 6.8037 - val_accuracy: 0.5805\n",
      "Epoch 44/50\n",
      "119997/119997 [==============================] - 4s 37us/sample - loss: 0.5417 - accuracy: 0.6689 - val_loss: 4.6639 - val_accuracy: 0.5243\n",
      "Epoch 45/50\n",
      "119997/119997 [==============================] - 4s 37us/sample - loss: 0.5420 - accuracy: 0.6657 - val_loss: 4.3621 - val_accuracy: 0.5113\n",
      "Epoch 46/50\n",
      "119997/119997 [==============================] - 4s 37us/sample - loss: 0.5214 - accuracy: 0.6875 - val_loss: 5.7658 - val_accuracy: 0.5297\n",
      "Epoch 47/50\n",
      "119997/119997 [==============================] - 4s 37us/sample - loss: 0.5252 - accuracy: 0.6862 - val_loss: 4.4341 - val_accuracy: 0.5297\n",
      "Epoch 48/50\n",
      "119997/119997 [==============================] - 4s 37us/sample - loss: 0.5407 - accuracy: 0.6752 - val_loss: 4.8822 - val_accuracy: 0.5372\n",
      "Epoch 49/50\n",
      "119997/119997 [==============================] - 5s 38us/sample - loss: 0.5309 - accuracy: 0.6766 - val_loss: 2.9532 - val_accuracy: 0.4966\n",
      "Epoch 50/50\n",
      "119997/119997 [==============================] - 4s 36us/sample - loss: 0.5188 - accuracy: 0.6906 - val_loss: 5.1675 - val_accuracy: 0.5470\n",
      "3999/3999 [==============================] - 0s 16us/sample - loss: 35.7075 - accuracy: 0.4046\n",
      "3999/3999 [==============================] - 0s 19us/sample - loss: 11.1926 - accuracy: 0.4979\n",
      "3999/3999 [==============================] - 0s 17us/sample - loss: 8.4911 - accuracy: 0.5246\n",
      "4-layers-32-nodes\n",
      "Train on 119997 samples, validate on 29997 samples\n",
      "Epoch 1/50\n",
      "119997/119997 [==============================] - 5s 41us/sample - loss: 0.6905 - accuracy: 0.5109 - val_loss: 0.6765 - val_accuracy: 0.5643\n",
      "Epoch 2/50\n",
      "119997/119997 [==============================] - 4s 37us/sample - loss: 0.6851 - accuracy: 0.5235 - val_loss: 0.6736 - val_accuracy: 0.5128\n",
      "Epoch 3/50\n",
      "119997/119997 [==============================] - 4s 36us/sample - loss: 0.6675 - accuracy: 0.5482 - val_loss: 1.0713 - val_accuracy: 0.5343\n",
      "Epoch 4/50\n",
      "119997/119997 [==============================] - 4s 36us/sample - loss: 0.6467 - accuracy: 0.5659 - val_loss: 0.7467 - val_accuracy: 0.4939\n",
      "Epoch 5/50\n",
      "119997/119997 [==============================] - 5s 39us/sample - loss: 0.6406 - accuracy: 0.5788 - val_loss: 1.1579 - val_accuracy: 0.5745\n",
      "Epoch 6/50\n",
      "119997/119997 [==============================] - 4s 37us/sample - loss: 0.5684 - accuracy: 0.6582 - val_loss: 1.2577 - val_accuracy: 0.5883\n",
      "Epoch 7/50\n",
      "119997/119997 [==============================] - 4s 36us/sample - loss: 0.5633 - accuracy: 0.6610 - val_loss: 1.3784 - val_accuracy: 0.6262\n",
      "Epoch 8/50\n",
      "119997/119997 [==============================] - 4s 37us/sample - loss: 0.5631 - accuracy: 0.6606 - val_loss: 1.2631 - val_accuracy: 0.6381\n",
      "Epoch 9/50\n",
      "119997/119997 [==============================] - 4s 37us/sample - loss: 0.5613 - accuracy: 0.6608 - val_loss: 1.2776 - val_accuracy: 0.5312\n",
      "Epoch 10/50\n",
      "119997/119997 [==============================] - 4s 36us/sample - loss: 0.5601 - accuracy: 0.6628 - val_loss: 1.4144 - val_accuracy: 0.5898\n",
      "Epoch 11/50\n",
      "119997/119997 [==============================] - 4s 37us/sample - loss: 0.5613 - accuracy: 0.6611 - val_loss: 1.3736 - val_accuracy: 0.6336\n",
      "Epoch 12/50\n",
      "119997/119997 [==============================] - 5s 38us/sample - loss: 0.5585 - accuracy: 0.6640 - val_loss: 1.3696 - val_accuracy: 0.5663\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 13/50\n",
      "119997/119997 [==============================] - 4s 37us/sample - loss: 0.5593 - accuracy: 0.6631 - val_loss: 1.5816 - val_accuracy: 0.5487\n",
      "Epoch 14/50\n",
      "119997/119997 [==============================] - 4s 37us/sample - loss: 0.5578 - accuracy: 0.6634 - val_loss: 1.7570 - val_accuracy: 0.6371\n",
      "Epoch 15/50\n",
      "119997/119997 [==============================] - 5s 38us/sample - loss: 0.5573 - accuracy: 0.6634 - val_loss: 1.6399 - val_accuracy: 0.6184\n",
      "Epoch 16/50\n",
      "119997/119997 [==============================] - 5s 39us/sample - loss: 0.5579 - accuracy: 0.6631 - val_loss: 1.6802 - val_accuracy: 0.6547\n",
      "Epoch 17/50\n",
      "119997/119997 [==============================] - 4s 37us/sample - loss: 0.5561 - accuracy: 0.6645 - val_loss: 1.6800 - val_accuracy: 0.6466\n",
      "Epoch 18/50\n",
      "119997/119997 [==============================] - 4s 37us/sample - loss: 0.5565 - accuracy: 0.6641 - val_loss: 1.3377 - val_accuracy: 0.6480\n",
      "Epoch 19/50\n",
      "119997/119997 [==============================] - 5s 38us/sample - loss: 0.5567 - accuracy: 0.6641 - val_loss: 1.5338 - val_accuracy: 0.6484\n",
      "Epoch 20/50\n",
      "119997/119997 [==============================] - 4s 37us/sample - loss: 0.5581 - accuracy: 0.6631 - val_loss: 1.6508 - val_accuracy: 0.6449\n",
      "Epoch 21/50\n",
      "119997/119997 [==============================] - 4s 37us/sample - loss: 0.5566 - accuracy: 0.6649 - val_loss: 1.5283 - val_accuracy: 0.6369\n",
      "Epoch 22/50\n",
      "119997/119997 [==============================] - 4s 37us/sample - loss: 0.5553 - accuracy: 0.6655 - val_loss: 1.8909 - val_accuracy: 0.5695\n",
      "Epoch 23/50\n",
      "119997/119997 [==============================] - 4s 37us/sample - loss: 0.5560 - accuracy: 0.6652 - val_loss: 2.2649 - val_accuracy: 0.6601\n",
      "Epoch 24/50\n",
      "119997/119997 [==============================] - 4s 37us/sample - loss: 0.5551 - accuracy: 0.6661 - val_loss: 2.1197 - val_accuracy: 0.6487\n",
      "Epoch 25/50\n",
      "119997/119997 [==============================] - 5s 38us/sample - loss: 0.5551 - accuracy: 0.6650 - val_loss: 2.0515 - val_accuracy: 0.6453\n",
      "Epoch 26/50\n",
      "119997/119997 [==============================] - 4s 37us/sample - loss: 0.5540 - accuracy: 0.6678 - val_loss: 2.4957 - val_accuracy: 0.6427\n",
      "Epoch 27/50\n",
      "119997/119997 [==============================] - 4s 37us/sample - loss: 0.5542 - accuracy: 0.6704 - val_loss: 2.0925 - val_accuracy: 0.6187\n",
      "Epoch 28/50\n",
      "119997/119997 [==============================] - 4s 37us/sample - loss: 0.5525 - accuracy: 0.6747 - val_loss: 2.2083 - val_accuracy: 0.6473\n",
      "Epoch 29/50\n",
      "119997/119997 [==============================] - 4s 37us/sample - loss: 0.5525 - accuracy: 0.6763 - val_loss: 2.2641 - val_accuracy: 0.6550\n",
      "Epoch 30/50\n",
      "119997/119997 [==============================] - 4s 37us/sample - loss: 0.5494 - accuracy: 0.6846 - val_loss: 2.1163 - val_accuracy: 0.6560\n",
      "Epoch 31/50\n",
      "119997/119997 [==============================] - 4s 37us/sample - loss: 0.5462 - accuracy: 0.6915 - val_loss: 2.0026 - val_accuracy: 0.6526\n",
      "Epoch 32/50\n",
      "119997/119997 [==============================] - 5s 39us/sample - loss: 0.5359 - accuracy: 0.7059 - val_loss: 2.1756 - val_accuracy: 0.6601\n",
      "Epoch 33/50\n",
      "119997/119997 [==============================] - 4s 37us/sample - loss: 0.5247 - accuracy: 0.7196 - val_loss: 2.4731 - val_accuracy: 0.6516\n",
      "Epoch 34/50\n",
      "119997/119997 [==============================] - 4s 37us/sample - loss: 0.5183 - accuracy: 0.7285 - val_loss: 2.2060 - val_accuracy: 0.6488\n",
      "Epoch 35/50\n",
      "119997/119997 [==============================] - 4s 37us/sample - loss: 0.5123 - accuracy: 0.7355 - val_loss: 2.5290 - val_accuracy: 0.6545\n",
      "Epoch 36/50\n",
      "119997/119997 [==============================] - 5s 38us/sample - loss: 0.5130 - accuracy: 0.7336 - val_loss: 2.5579 - val_accuracy: 0.6524\n",
      "Epoch 37/50\n",
      "119997/119997 [==============================] - 5s 38us/sample - loss: 0.5078 - accuracy: 0.7371 - val_loss: 2.4383 - val_accuracy: 0.6302\n",
      "Epoch 38/50\n",
      "119997/119997 [==============================] - 5s 38us/sample - loss: 0.5047 - accuracy: 0.7390 - val_loss: 2.5629 - val_accuracy: 0.6343\n",
      "Epoch 39/50\n",
      "119997/119997 [==============================] - 5s 38us/sample - loss: 0.5108 - accuracy: 0.7342 - val_loss: 2.6653 - val_accuracy: 0.6468\n",
      "Epoch 40/50\n",
      "119997/119997 [==============================] - 5s 38us/sample - loss: 0.5078 - accuracy: 0.7365 - val_loss: 2.4207 - val_accuracy: 0.6525\n",
      "Epoch 41/50\n",
      "119997/119997 [==============================] - 5s 38us/sample - loss: 0.5056 - accuracy: 0.7401 - val_loss: 2.3960 - val_accuracy: 0.6682\n",
      "Epoch 42/50\n",
      "119997/119997 [==============================] - 5s 40us/sample - loss: 0.5090 - accuracy: 0.7373 - val_loss: 2.5085 - val_accuracy: 0.6416\n",
      "Epoch 43/50\n",
      "119997/119997 [==============================] - 4s 37us/sample - loss: 0.5059 - accuracy: 0.7391 - val_loss: 2.6849 - val_accuracy: 0.6461\n",
      "Epoch 44/50\n",
      "119997/119997 [==============================] - 4s 37us/sample - loss: 0.5079 - accuracy: 0.7392 - val_loss: 2.7041 - val_accuracy: 0.6436\n",
      "Epoch 45/50\n",
      "119997/119997 [==============================] - 4s 37us/sample - loss: 0.5117 - accuracy: 0.7337 - val_loss: 2.5348 - val_accuracy: 0.6434\n",
      "Epoch 46/50\n",
      "119997/119997 [==============================] - 4s 37us/sample - loss: 0.5140 - accuracy: 0.7280 - val_loss: 2.6295 - val_accuracy: 0.6485\n",
      "Epoch 47/50\n",
      "119997/119997 [==============================] - 4s 37us/sample - loss: 0.5045 - accuracy: 0.7392 - val_loss: 2.5574 - val_accuracy: 0.6526\n",
      "Epoch 48/50\n",
      "119997/119997 [==============================] - 5s 38us/sample - loss: 0.4907 - accuracy: 0.7496 - val_loss: 2.8094 - val_accuracy: 0.6551\n",
      "Epoch 49/50\n",
      "119997/119997 [==============================] - 4s 37us/sample - loss: 0.4930 - accuracy: 0.7511 - val_loss: 2.6615 - val_accuracy: 0.6448\n",
      "Epoch 50/50\n",
      "119997/119997 [==============================] - 4s 37us/sample - loss: 0.4870 - accuracy: 0.7544 - val_loss: 2.4847 - val_accuracy: 0.6938\n",
      "3999/3999 [==============================] - 0s 15us/sample - loss: 6.7238 - accuracy: 0.4484\n",
      "3999/3999 [==============================] - 0s 15us/sample - loss: 2.4517 - accuracy: 0.4764\n",
      "3999/3999 [==============================] - 0s 17us/sample - loss: 2.2636 - accuracy: 0.4966\n",
      "4-layers-64-nodes\n",
      "Train on 119997 samples, validate on 29997 samples\n",
      "Epoch 1/50\n",
      "119997/119997 [==============================] - 5s 43us/sample - loss: 0.6904 - accuracy: 0.5064 - val_loss: 0.6863 - val_accuracy: 0.4826\n",
      "Epoch 2/50\n",
      "119997/119997 [==============================] - 5s 41us/sample - loss: 0.6873 - accuracy: 0.5155 - val_loss: 0.6947 - val_accuracy: 0.4772\n",
      "Epoch 3/50\n",
      "119997/119997 [==============================] - 5s 38us/sample - loss: 0.6822 - accuracy: 0.5210 - val_loss: 0.8387 - val_accuracy: 0.4948\n",
      "Epoch 4/50\n",
      "119997/119997 [==============================] - 5s 38us/sample - loss: 0.6657 - accuracy: 0.5425 - val_loss: 1.0720 - val_accuracy: 0.4902\n",
      "Epoch 5/50\n",
      "119997/119997 [==============================] - 5s 38us/sample - loss: 0.6807 - accuracy: 0.5222 - val_loss: 0.7768 - val_accuracy: 0.4923\n",
      "Epoch 6/50\n",
      "119997/119997 [==============================] - 5s 38us/sample - loss: 0.6784 - accuracy: 0.5276 - val_loss: 1.3921 - val_accuracy: 0.5009\n",
      "Epoch 7/50\n",
      "119997/119997 [==============================] - 5s 38us/sample - loss: 0.6120 - accuracy: 0.6195 - val_loss: 2.7742 - val_accuracy: 0.5176\n",
      "Epoch 8/50\n",
      "119997/119997 [==============================] - 5s 40us/sample - loss: 0.5305 - accuracy: 0.7075 - val_loss: 4.4720 - val_accuracy: 0.5160\n",
      "Epoch 9/50\n",
      "119997/119997 [==============================] - 5s 39us/sample - loss: 0.5226 - accuracy: 0.7094 - val_loss: 5.1624 - val_accuracy: 0.5138\n",
      "Epoch 10/50\n",
      "119997/119997 [==============================] - 5s 38us/sample - loss: 0.5203 - accuracy: 0.7114 - val_loss: 5.7223 - val_accuracy: 0.5117\n",
      "Epoch 11/50\n",
      "119997/119997 [==============================] - 5s 38us/sample - loss: 0.5201 - accuracy: 0.7097 - val_loss: 5.9550 - val_accuracy: 0.5103\n",
      "Epoch 12/50\n",
      "119997/119997 [==============================] - 5s 38us/sample - loss: 0.5170 - accuracy: 0.7124 - val_loss: 6.5950 - val_accuracy: 0.5037\n",
      "Epoch 13/50\n",
      "119997/119997 [==============================] - 5s 38us/sample - loss: 0.5124 - accuracy: 0.7165 - val_loss: 6.5087 - val_accuracy: 0.5097\n",
      "Epoch 14/50\n",
      "119997/119997 [==============================] - 5s 38us/sample - loss: 0.5123 - accuracy: 0.7194 - val_loss: 6.3566 - val_accuracy: 0.5202\n",
      "Epoch 15/50\n",
      "119997/119997 [==============================] - 5s 40us/sample - loss: 0.5133 - accuracy: 0.7185 - val_loss: 6.0772 - val_accuracy: 0.5113\n",
      "Epoch 16/50\n",
      "119997/119997 [==============================] - 5s 38us/sample - loss: 0.5091 - accuracy: 0.7198 - val_loss: 6.7418 - val_accuracy: 0.5045\n",
      "Epoch 17/50\n",
      "119997/119997 [==============================] - 5s 38us/sample - loss: 0.5088 - accuracy: 0.7163 - val_loss: 7.0044 - val_accuracy: 0.5014\n",
      "Epoch 18/50\n",
      "119997/119997 [==============================] - 5s 41us/sample - loss: 0.5102 - accuracy: 0.7200 - val_loss: 6.5216 - val_accuracy: 0.5035\n",
      "Epoch 19/50\n",
      "119997/119997 [==============================] - 5s 38us/sample - loss: 0.5109 - accuracy: 0.7197 - val_loss: 6.1266 - val_accuracy: 0.4990\n",
      "Epoch 20/50\n",
      "119997/119997 [==============================] - 5s 38us/sample - loss: 0.5091 - accuracy: 0.7168 - val_loss: 6.6518 - val_accuracy: 0.5024\n",
      "Epoch 21/50\n",
      "119997/119997 [==============================] - 5s 40us/sample - loss: 0.5076 - accuracy: 0.7209 - val_loss: 6.4439 - val_accuracy: 0.5070\n",
      "Epoch 22/50\n",
      "119997/119997 [==============================] - 5s 38us/sample - loss: 0.5036 - accuracy: 0.7253 - val_loss: 5.8795 - val_accuracy: 0.4990\n",
      "Epoch 23/50\n",
      "119997/119997 [==============================] - 5s 38us/sample - loss: 0.5058 - accuracy: 0.7210 - val_loss: 6.0713 - val_accuracy: 0.5044\n",
      "Epoch 24/50\n",
      "119997/119997 [==============================] - 5s 39us/sample - loss: 0.5060 - accuracy: 0.7207 - val_loss: 6.7950 - val_accuracy: 0.5051\n",
      "Epoch 25/50\n",
      "119997/119997 [==============================] - 5s 38us/sample - loss: 0.5100 - accuracy: 0.7214 - val_loss: 6.2977 - val_accuracy: 0.5063\n",
      "Epoch 26/50\n",
      "119997/119997 [==============================] - 5s 38us/sample - loss: 0.4911 - accuracy: 0.7435 - val_loss: 5.8103 - val_accuracy: 0.4954\n",
      "Epoch 27/50\n",
      "119997/119997 [==============================] - 5s 38us/sample - loss: 0.4947 - accuracy: 0.7327 - val_loss: 6.9913 - val_accuracy: 0.5070\n",
      "Epoch 28/50\n",
      "119997/119997 [==============================] - 5s 40us/sample - loss: 0.5026 - accuracy: 0.7231 - val_loss: 6.0399 - val_accuracy: 0.4987\n",
      "Epoch 29/50\n",
      "119997/119997 [==============================] - 5s 38us/sample - loss: 0.4985 - accuracy: 0.7290 - val_loss: 6.4504 - val_accuracy: 0.4996\n",
      "Epoch 30/50\n",
      "119997/119997 [==============================] - 5s 38us/sample - loss: 0.4958 - accuracy: 0.7337 - val_loss: 6.3397 - val_accuracy: 0.5010\n",
      "Epoch 31/50\n",
      "119997/119997 [==============================] - 5s 39us/sample - loss: 0.4926 - accuracy: 0.7365 - val_loss: 8.0123 - val_accuracy: 0.5027\n",
      "Epoch 32/50\n",
      "119997/119997 [==============================] - 5s 38us/sample - loss: 0.4919 - accuracy: 0.7380 - val_loss: 7.0765 - val_accuracy: 0.5023\n",
      "Epoch 33/50\n",
      "119997/119997 [==============================] - 5s 38us/sample - loss: 0.4964 - accuracy: 0.7295 - val_loss: 7.6317 - val_accuracy: 0.5018\n",
      "Epoch 34/50\n",
      "119997/119997 [==============================] - 5s 39us/sample - loss: 0.4950 - accuracy: 0.7333 - val_loss: 7.5563 - val_accuracy: 0.5008\n",
      "Epoch 35/50\n",
      "119997/119997 [==============================] - 5s 40us/sample - loss: 0.5001 - accuracy: 0.7287 - val_loss: 4.6499 - val_accuracy: 0.4953\n",
      "Epoch 36/50\n",
      "119997/119997 [==============================] - 5s 38us/sample - loss: 0.4907 - accuracy: 0.7379 - val_loss: 6.5027 - val_accuracy: 0.4978\n",
      "Epoch 37/50\n",
      "119997/119997 [==============================] - 5s 38us/sample - loss: 0.4932 - accuracy: 0.7350 - val_loss: 8.7899 - val_accuracy: 0.5021\n",
      "Epoch 38/50\n",
      "119997/119997 [==============================] - 5s 38us/sample - loss: 0.4988 - accuracy: 0.7279 - val_loss: 8.6125 - val_accuracy: 0.4988\n",
      "Epoch 39/50\n",
      "119997/119997 [==============================] - 5s 38us/sample - loss: 0.5015 - accuracy: 0.7254 - val_loss: 6.5212 - val_accuracy: 0.5017\n",
      "Epoch 40/50\n",
      "119997/119997 [==============================] - 5s 38us/sample - loss: 0.4877 - accuracy: 0.7422 - val_loss: 8.2636 - val_accuracy: 0.5062\n",
      "Epoch 41/50\n",
      "119997/119997 [==============================] - 5s 39us/sample - loss: 0.4906 - accuracy: 0.7369 - val_loss: 8.8153 - val_accuracy: 0.5019\n",
      "Epoch 42/50\n",
      "119997/119997 [==============================] - 5s 38us/sample - loss: 0.4946 - accuracy: 0.7316 - val_loss: 6.9999 - val_accuracy: 0.4979\n",
      "Epoch 43/50\n",
      "119997/119997 [==============================] - 5s 38us/sample - loss: 0.4925 - accuracy: 0.7359 - val_loss: 7.1743 - val_accuracy: 0.5006\n",
      "Epoch 44/50\n",
      "119997/119997 [==============================] - 5s 41us/sample - loss: 0.4934 - accuracy: 0.7324 - val_loss: 8.2559 - val_accuracy: 0.4980\n",
      "Epoch 45/50\n",
      "119997/119997 [==============================] - 5s 39us/sample - loss: 0.4904 - accuracy: 0.7372 - val_loss: 6.4977 - val_accuracy: 0.4959\n",
      "Epoch 46/50\n",
      "119997/119997 [==============================] - 5s 38us/sample - loss: 0.4920 - accuracy: 0.7349 - val_loss: 8.3210 - val_accuracy: 0.5086\n",
      "Epoch 47/50\n",
      "119997/119997 [==============================] - 5s 40us/sample - loss: 0.4856 - accuracy: 0.7431 - val_loss: 9.1060 - val_accuracy: 0.5065\n",
      "Epoch 48/50\n",
      "119997/119997 [==============================] - 5s 39us/sample - loss: 0.4836 - accuracy: 0.7427 - val_loss: 8.6468 - val_accuracy: 0.5008\n",
      "Epoch 49/50\n",
      "119997/119997 [==============================] - 5s 38us/sample - loss: 0.4833 - accuracy: 0.7454 - val_loss: 9.6455 - val_accuracy: 0.4992\n",
      "Epoch 50/50\n",
      "119997/119997 [==============================] - 5s 38us/sample - loss: 0.4926 - accuracy: 0.7332 - val_loss: 9.4719 - val_accuracy: 0.4975\n",
      "3999/3999 [==============================] - 0s 20us/sample - loss: 3.8423 - accuracy: 0.5731\n",
      "3999/3999 [==============================] - 0s 18us/sample - loss: 5.7195 - accuracy: 0.5754\n",
      "3999/3999 [==============================] - 0s 20us/sample - loss: 3.9733 - accuracy: 0.6657\n",
      "4-layers-128-nodes\n",
      "Train on 119997 samples, validate on 29997 samples\n",
      "Epoch 1/50\n",
      "119997/119997 [==============================] - 6s 49us/sample - loss: 0.6901 - accuracy: 0.5073 - val_loss: 0.6794 - val_accuracy: 0.5466\n",
      "Epoch 2/50\n",
      "119997/119997 [==============================] - 5s 44us/sample - loss: 0.6842 - accuracy: 0.5214 - val_loss: 0.6997 - val_accuracy: 0.5166\n",
      "Epoch 3/50\n",
      "119997/119997 [==============================] - 6s 46us/sample - loss: 0.6775 - accuracy: 0.5297 - val_loss: 0.8409 - val_accuracy: 0.5487\n",
      "Epoch 4/50\n",
      "119997/119997 [==============================] - 5s 43us/sample - loss: 0.6718 - accuracy: 0.5329 - val_loss: 0.9697 - val_accuracy: 0.5458\n",
      "Epoch 5/50\n",
      "119997/119997 [==============================] - 5s 44us/sample - loss: 0.6723 - accuracy: 0.5296 - val_loss: 0.8521 - val_accuracy: 0.5497\n",
      "Epoch 6/50\n",
      "119997/119997 [==============================] - 5s 44us/sample - loss: 0.6741 - accuracy: 0.5249 - val_loss: 1.0137 - val_accuracy: 0.5284\n",
      "Epoch 7/50\n",
      "119997/119997 [==============================] - 5s 44us/sample - loss: 0.6729 - accuracy: 0.5301 - val_loss: 0.9273 - val_accuracy: 0.5480\n",
      "Epoch 8/50\n",
      "119997/119997 [==============================] - 5s 45us/sample - loss: 0.6726 - accuracy: 0.5308 - val_loss: 0.8473 - val_accuracy: 0.5373\n",
      "Epoch 9/50\n",
      "119997/119997 [==============================] - 5s 44us/sample - loss: 0.6776 - accuracy: 0.5263 - val_loss: 0.9209 - val_accuracy: 0.5034\n",
      "Epoch 10/50\n",
      "119997/119997 [==============================] - 5s 43us/sample - loss: 0.6745 - accuracy: 0.5327 - val_loss: 0.9170 - val_accuracy: 0.5406\n",
      "Epoch 11/50\n",
      "119997/119997 [==============================] - 5s 44us/sample - loss: 0.6684 - accuracy: 0.5492 - val_loss: 0.8691 - val_accuracy: 0.5381\n",
      "Epoch 12/50\n",
      "119997/119997 [==============================] - 5s 44us/sample - loss: 0.6246 - accuracy: 0.6104 - val_loss: 1.1132 - val_accuracy: 0.5782\n",
      "Epoch 13/50\n",
      "119997/119997 [==============================] - 5s 43us/sample - loss: 0.5624 - accuracy: 0.6658 - val_loss: 1.4037 - val_accuracy: 0.5705\n",
      "Epoch 14/50\n",
      "119997/119997 [==============================] - 5s 46us/sample - loss: 0.5525 - accuracy: 0.6701 - val_loss: 1.3431 - val_accuracy: 0.6252\n",
      "Epoch 15/50\n",
      "119997/119997 [==============================] - 5s 44us/sample - loss: 0.5488 - accuracy: 0.6733 - val_loss: 1.5724 - val_accuracy: 0.6237\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 16/50\n",
      "119997/119997 [==============================] - 5s 44us/sample - loss: 0.5495 - accuracy: 0.6716 - val_loss: 1.4729 - val_accuracy: 0.6330\n",
      "Epoch 17/50\n",
      "119997/119997 [==============================] - 6s 47us/sample - loss: 0.5487 - accuracy: 0.6715 - val_loss: 1.2620 - val_accuracy: 0.5983\n",
      "Epoch 18/50\n",
      "119997/119997 [==============================] - 5s 44us/sample - loss: 0.5484 - accuracy: 0.6712 - val_loss: 1.6114 - val_accuracy: 0.6270\n",
      "Epoch 19/50\n",
      "119997/119997 [==============================] - 5s 44us/sample - loss: 0.5465 - accuracy: 0.6718 - val_loss: 1.3967 - val_accuracy: 0.6327\n",
      "Epoch 20/50\n",
      "119997/119997 [==============================] - 6s 49us/sample - loss: 0.5464 - accuracy: 0.6722 - val_loss: 1.2839 - val_accuracy: 0.6446\n",
      "Epoch 21/50\n",
      "119997/119997 [==============================] - 5s 45us/sample - loss: 0.5426 - accuracy: 0.6789 - val_loss: 1.3084 - val_accuracy: 0.6138\n",
      "Epoch 22/50\n",
      "119997/119997 [==============================] - 6s 49us/sample - loss: 0.5309 - accuracy: 0.6976 - val_loss: 1.4052 - val_accuracy: 0.6150\n",
      "Epoch 23/50\n",
      "119997/119997 [==============================] - 6s 52us/sample - loss: 0.5163 - accuracy: 0.7113 - val_loss: 1.1949 - val_accuracy: 0.6228\n",
      "Epoch 24/50\n",
      "119997/119997 [==============================] - 6s 47us/sample - loss: 0.5108 - accuracy: 0.7152 - val_loss: 1.3197 - val_accuracy: 0.6479\n",
      "Epoch 25/50\n",
      "119997/119997 [==============================] - 5s 44us/sample - loss: 0.5096 - accuracy: 0.7160 - val_loss: 1.5532 - val_accuracy: 0.6279\n",
      "Epoch 26/50\n",
      "119997/119997 [==============================] - 5s 46us/sample - loss: 0.5068 - accuracy: 0.7155 - val_loss: 1.0554 - val_accuracy: 0.6605\n",
      "Epoch 27/50\n",
      "119997/119997 [==============================] - 5s 43us/sample - loss: 0.5058 - accuracy: 0.7175 - val_loss: 1.2769 - val_accuracy: 0.6505\n",
      "Epoch 28/50\n",
      "119997/119997 [==============================] - 6s 46us/sample - loss: 0.5072 - accuracy: 0.7159 - val_loss: 1.7147 - val_accuracy: 0.6027\n",
      "Epoch 29/50\n",
      "119997/119997 [==============================] - 5s 44us/sample - loss: 0.5085 - accuracy: 0.7139 - val_loss: 1.2522 - val_accuracy: 0.6357\n",
      "Epoch 30/50\n",
      "119997/119997 [==============================] - 6s 46us/sample - loss: 0.5064 - accuracy: 0.7153 - val_loss: 1.6516 - val_accuracy: 0.6362\n",
      "Epoch 31/50\n",
      "119997/119997 [==============================] - 6s 47us/sample - loss: 0.5053 - accuracy: 0.7167 - val_loss: 1.1400 - val_accuracy: 0.6674\n",
      "Epoch 32/50\n",
      "119997/119997 [==============================] - 5s 44us/sample - loss: 0.5039 - accuracy: 0.7174 - val_loss: 1.4896 - val_accuracy: 0.6664\n",
      "Epoch 33/50\n",
      "119997/119997 [==============================] - 6s 48us/sample - loss: 0.5028 - accuracy: 0.7199 - val_loss: 1.5536 - val_accuracy: 0.6415\n",
      "Epoch 34/50\n",
      "119997/119997 [==============================] - 5s 43us/sample - loss: 0.5049 - accuracy: 0.7153 - val_loss: 1.2060 - val_accuracy: 0.6533\n",
      "Epoch 35/50\n",
      "119997/119997 [==============================] - 5s 42us/sample - loss: 0.5046 - accuracy: 0.7168 - val_loss: 1.7483 - val_accuracy: 0.6553\n",
      "Epoch 36/50\n",
      "119997/119997 [==============================] - 5s 46us/sample - loss: 0.5031 - accuracy: 0.7171 - val_loss: 1.3138 - val_accuracy: 0.6880\n",
      "Epoch 37/50\n",
      "119997/119997 [==============================] - 5s 43us/sample - loss: 0.5009 - accuracy: 0.7224 - val_loss: 1.4035 - val_accuracy: 0.6558\n",
      "Epoch 38/50\n",
      "119997/119997 [==============================] - 5s 43us/sample - loss: 0.4961 - accuracy: 0.7279 - val_loss: 1.7110 - val_accuracy: 0.6389\n",
      "Epoch 39/50\n",
      "119997/119997 [==============================] - 6s 46us/sample - loss: 0.4993 - accuracy: 0.7242 - val_loss: 1.3837 - val_accuracy: 0.6231\n",
      "Epoch 40/50\n",
      "119997/119997 [==============================] - 6s 46us/sample - loss: 0.4975 - accuracy: 0.7192 - val_loss: 2.1564 - val_accuracy: 0.6077\n",
      "Epoch 41/50\n",
      "119997/119997 [==============================] - 5s 45us/sample - loss: 0.5023 - accuracy: 0.7142 - val_loss: 1.4846 - val_accuracy: 0.6095\n",
      "Epoch 42/50\n",
      "119997/119997 [==============================] - 5s 45us/sample - loss: 0.5041 - accuracy: 0.7172 - val_loss: 2.3095 - val_accuracy: 0.5910\n",
      "Epoch 43/50\n",
      "119997/119997 [==============================] - 5s 44us/sample - loss: 0.5019 - accuracy: 0.7191 - val_loss: 1.7825 - val_accuracy: 0.5980\n",
      "Epoch 44/50\n",
      "119997/119997 [==============================] - 5s 44us/sample - loss: 0.4980 - accuracy: 0.7234 - val_loss: 1.7946 - val_accuracy: 0.6145\n",
      "Epoch 45/50\n",
      "119997/119997 [==============================] - 5s 44us/sample - loss: 0.4956 - accuracy: 0.7186 - val_loss: 2.1727 - val_accuracy: 0.5966\n",
      "Epoch 46/50\n",
      "119997/119997 [==============================] - 5s 43us/sample - loss: 0.4955 - accuracy: 0.7155 - val_loss: 2.8126 - val_accuracy: 0.5865\n",
      "Epoch 47/50\n",
      "119997/119997 [==============================] - 5s 44us/sample - loss: 0.5047 - accuracy: 0.7145 - val_loss: 2.2858 - val_accuracy: 0.5845\n",
      "Epoch 48/50\n",
      "119997/119997 [==============================] - 5s 43us/sample - loss: 0.4993 - accuracy: 0.7237 - val_loss: 2.3246 - val_accuracy: 0.6252\n",
      "Epoch 49/50\n",
      "119997/119997 [==============================] - 5s 43us/sample - loss: 0.5029 - accuracy: 0.7179 - val_loss: 2.8096 - val_accuracy: 0.6163\n",
      "Epoch 50/50\n",
      "119997/119997 [==============================] - 5s 43us/sample - loss: 0.4951 - accuracy: 0.7302 - val_loss: 3.6806 - val_accuracy: 0.6082\n",
      "3999/3999 [==============================] - 0s 17us/sample - loss: 7.0251 - accuracy: 0.4861\n",
      "3999/3999 [==============================] - 0s 21us/sample - loss: 5.5317 - accuracy: 0.7009\n",
      "3999/3999 [==============================] - 0s 20us/sample - loss: 13.0820 - accuracy: 0.7072\n",
      "5-layers-16-nodes\n",
      "Train on 119997 samples, validate on 29997 samples\n",
      "Epoch 1/50\n",
      "119997/119997 [==============================] - 5s 42us/sample - loss: 0.6910 - accuracy: 0.5072 - val_loss: 0.6831 - val_accuracy: 0.5292\n",
      "Epoch 2/50\n",
      "119997/119997 [==============================] - 5s 39us/sample - loss: 0.6884 - accuracy: 0.5139 - val_loss: 0.7076 - val_accuracy: 0.5234\n",
      "Epoch 3/50\n",
      "119997/119997 [==============================] - 5s 41us/sample - loss: 0.6830 - accuracy: 0.5225 - val_loss: 0.7228 - val_accuracy: 0.4875\n",
      "Epoch 4/50\n",
      "119997/119997 [==============================] - 5s 40us/sample - loss: 0.6635 - accuracy: 0.5493 - val_loss: 0.9657 - val_accuracy: 0.5563\n",
      "Epoch 5/50\n",
      "119997/119997 [==============================] - 5s 39us/sample - loss: 0.6625 - accuracy: 0.5445 - val_loss: 1.0366 - val_accuracy: 0.4747\n",
      "Epoch 6/50\n",
      "119997/119997 [==============================] - 5s 40us/sample - loss: 0.6341 - accuracy: 0.5936 - val_loss: 1.1398 - val_accuracy: 0.5522\n",
      "Epoch 7/50\n",
      "119997/119997 [==============================] - 5s 38us/sample - loss: 0.6532 - accuracy: 0.5735 - val_loss: 1.2602 - val_accuracy: 0.4662\n",
      "Epoch 8/50\n",
      "119997/119997 [==============================] - 5s 38us/sample - loss: 0.6504 - accuracy: 0.5572 - val_loss: 1.0998 - val_accuracy: 0.4673\n",
      "Epoch 9/50\n",
      "119997/119997 [==============================] - 5s 38us/sample - loss: 0.6605 - accuracy: 0.5429 - val_loss: 0.9761 - val_accuracy: 0.4909\n",
      "Epoch 10/50\n",
      "119997/119997 [==============================] - 5s 43us/sample - loss: 0.6471 - accuracy: 0.5605 - val_loss: 0.8920 - val_accuracy: 0.4873\n",
      "Epoch 11/50\n",
      "119997/119997 [==============================] - 5s 41us/sample - loss: 0.6546 - accuracy: 0.5485 - val_loss: 0.8037 - val_accuracy: 0.4878\n",
      "Epoch 12/50\n",
      "119997/119997 [==============================] - 5s 40us/sample - loss: 0.6425 - accuracy: 0.5671 - val_loss: 0.9531 - val_accuracy: 0.4921\n",
      "Epoch 13/50\n",
      "119997/119997 [==============================] - 5s 40us/sample - loss: 0.6645 - accuracy: 0.5399 - val_loss: 0.8515 - val_accuracy: 0.4736\n",
      "Epoch 14/50\n",
      "119997/119997 [==============================] - 5s 40us/sample - loss: 0.6658 - accuracy: 0.5367 - val_loss: 0.9292 - val_accuracy: 0.5312\n",
      "Epoch 15/50\n",
      "119997/119997 [==============================] - 5s 40us/sample - loss: 0.6456 - accuracy: 0.5676 - val_loss: 0.8673 - val_accuracy: 0.4818\n",
      "Epoch 16/50\n",
      "119997/119997 [==============================] - 5s 40us/sample - loss: 0.6550 - accuracy: 0.5524 - val_loss: 1.2161 - val_accuracy: 0.4945\n",
      "Epoch 17/50\n",
      "119997/119997 [==============================] - 5s 39us/sample - loss: 0.6365 - accuracy: 0.5778 - val_loss: 1.5202 - val_accuracy: 0.4923\n",
      "Epoch 18/50\n",
      "119997/119997 [==============================] - 5s 38us/sample - loss: 0.6361 - accuracy: 0.5826 - val_loss: 1.2219 - val_accuracy: 0.5280\n",
      "Epoch 19/50\n",
      "119997/119997 [==============================] - 5s 38us/sample - loss: 0.6544 - accuracy: 0.5524 - val_loss: 1.3426 - val_accuracy: 0.4887\n",
      "Epoch 20/50\n",
      "119997/119997 [==============================] - 5s 38us/sample - loss: 0.6688 - accuracy: 0.5325 - val_loss: 1.0235 - val_accuracy: 0.4844\n",
      "Epoch 21/50\n",
      "119997/119997 [==============================] - 5s 40us/sample - loss: 0.6358 - accuracy: 0.5842 - val_loss: 1.6467 - val_accuracy: 0.4973\n",
      "Epoch 22/50\n",
      "119997/119997 [==============================] - 5s 43us/sample - loss: 0.6342 - accuracy: 0.5974 - val_loss: 1.0778 - val_accuracy: 0.5352\n",
      "Epoch 23/50\n",
      "119997/119997 [==============================] - 5s 38us/sample - loss: 0.6227 - accuracy: 0.6183 - val_loss: 1.7044 - val_accuracy: 0.4967\n",
      "Epoch 24/50\n",
      "119997/119997 [==============================] - 5s 38us/sample - loss: 0.5931 - accuracy: 0.6636 - val_loss: 2.1824 - val_accuracy: 0.5012\n",
      "Epoch 25/50\n",
      "119997/119997 [==============================] - 5s 39us/sample - loss: 0.5635 - accuracy: 0.6621 - val_loss: 3.1637 - val_accuracy: 0.5057\n",
      "Epoch 26/50\n",
      "119997/119997 [==============================] - 5s 39us/sample - loss: 0.5631 - accuracy: 0.6623 - val_loss: 3.2325 - val_accuracy: 0.5013\n",
      "Epoch 27/50\n",
      "119997/119997 [==============================] - 5s 40us/sample - loss: 0.5585 - accuracy: 0.6698 - val_loss: 3.1303 - val_accuracy: 0.5097\n",
      "Epoch 28/50\n",
      "119997/119997 [==============================] - 5s 40us/sample - loss: 0.5563 - accuracy: 0.6757 - val_loss: 3.2549 - val_accuracy: 0.5140\n",
      "Epoch 29/50\n",
      "119997/119997 [==============================] - 5s 39us/sample - loss: 0.5550 - accuracy: 0.6795 - val_loss: 3.5043 - val_accuracy: 0.5348\n",
      "Epoch 30/50\n",
      "119997/119997 [==============================] - 5s 39us/sample - loss: 0.5552 - accuracy: 0.6793 - val_loss: 3.4778 - val_accuracy: 0.5087\n",
      "Epoch 31/50\n",
      "119997/119997 [==============================] - 5s 38us/sample - loss: 0.5495 - accuracy: 0.6845 - val_loss: 3.6935 - val_accuracy: 0.5179\n",
      "Epoch 32/50\n",
      "119997/119997 [==============================] - 5s 38us/sample - loss: 0.5464 - accuracy: 0.6829 - val_loss: 3.7190 - val_accuracy: 0.5372\n",
      "Epoch 33/50\n",
      "119997/119997 [==============================] - 5s 40us/sample - loss: 0.5418 - accuracy: 0.6932 - val_loss: 3.9039 - val_accuracy: 0.5347\n",
      "Epoch 34/50\n",
      "119997/119997 [==============================] - 5s 38us/sample - loss: 0.5291 - accuracy: 0.7078 - val_loss: 3.7334 - val_accuracy: 0.4972\n",
      "Epoch 35/50\n",
      "119997/119997 [==============================] - 5s 40us/sample - loss: 0.5266 - accuracy: 0.7026 - val_loss: 3.9547 - val_accuracy: 0.5408\n",
      "Epoch 36/50\n",
      "119997/119997 [==============================] - 5s 38us/sample - loss: 0.5267 - accuracy: 0.7090 - val_loss: 3.5321 - val_accuracy: 0.5446\n",
      "Epoch 37/50\n",
      "119997/119997 [==============================] - 5s 38us/sample - loss: 0.5079 - accuracy: 0.7378 - val_loss: 4.1284 - val_accuracy: 0.5332\n",
      "Epoch 38/50\n",
      "119997/119997 [==============================] - 5s 40us/sample - loss: 0.5120 - accuracy: 0.7427 - val_loss: 3.9868 - val_accuracy: 0.5404\n",
      "Epoch 39/50\n",
      "119997/119997 [==============================] - 5s 38us/sample - loss: 0.4936 - accuracy: 0.7553 - val_loss: 3.5473 - val_accuracy: 0.5213\n",
      "Epoch 40/50\n",
      "119997/119997 [==============================] - 5s 38us/sample - loss: 0.4755 - accuracy: 0.7642 - val_loss: 3.9571 - val_accuracy: 0.5189\n",
      "Epoch 41/50\n",
      "119997/119997 [==============================] - 5s 39us/sample - loss: 0.4661 - accuracy: 0.7677 - val_loss: 3.7658 - val_accuracy: 0.5002\n",
      "Epoch 42/50\n",
      "119997/119997 [==============================] - 5s 39us/sample - loss: 0.4628 - accuracy: 0.7681 - val_loss: 3.7972 - val_accuracy: 0.5153\n",
      "Epoch 43/50\n",
      "119997/119997 [==============================] - 5s 38us/sample - loss: 0.4553 - accuracy: 0.7721 - val_loss: 3.6582 - val_accuracy: 0.5396\n",
      "Epoch 44/50\n",
      "119997/119997 [==============================] - 5s 38us/sample - loss: 0.4499 - accuracy: 0.7748 - val_loss: 3.5116 - val_accuracy: 0.5498\n",
      "Epoch 45/50\n",
      "119997/119997 [==============================] - 5s 38us/sample - loss: 0.4470 - accuracy: 0.7784 - val_loss: 3.9083 - val_accuracy: 0.5118\n",
      "Epoch 46/50\n",
      "119997/119997 [==============================] - 5s 38us/sample - loss: 0.4569 - accuracy: 0.7730 - val_loss: 4.0418 - val_accuracy: 0.4926\n",
      "Epoch 47/50\n",
      "119997/119997 [==============================] - 5s 38us/sample - loss: 0.4563 - accuracy: 0.7716 - val_loss: 4.3266 - val_accuracy: 0.5192\n",
      "Epoch 48/50\n",
      "119997/119997 [==============================] - 5s 39us/sample - loss: 0.4513 - accuracy: 0.7744 - val_loss: 3.6308 - val_accuracy: 0.5313\n",
      "Epoch 49/50\n",
      "119997/119997 [==============================] - 5s 40us/sample - loss: 0.4425 - accuracy: 0.7799 - val_loss: 3.5390 - val_accuracy: 0.5483\n",
      "Epoch 50/50\n",
      "119997/119997 [==============================] - 5s 38us/sample - loss: 0.4356 - accuracy: 0.7836 - val_loss: 3.6542 - val_accuracy: 0.5340\n",
      "3999/3999 [==============================] - 0s 20us/sample - loss: 6.4445 - accuracy: 0.4509\n",
      "3999/3999 [==============================] - 0s 17us/sample - loss: 2.7136 - accuracy: 0.5781\n",
      "3999/3999 [==============================] - 0s 17us/sample - loss: 17.6830 - accuracy: 0.2291\n",
      "5-layers-32-nodes\n",
      "Train on 119997 samples, validate on 29997 samples\n",
      "Epoch 1/50\n",
      "119997/119997 [==============================] - 5s 42us/sample - loss: 0.6905 - accuracy: 0.5061 - val_loss: 0.6704 - val_accuracy: 0.4739\n",
      "Epoch 2/50\n",
      "119997/119997 [==============================] - 5s 38us/sample - loss: 0.6874 - accuracy: 0.5172 - val_loss: 0.6696 - val_accuracy: 0.5084\n",
      "Epoch 3/50\n",
      "119997/119997 [==============================] - 5s 38us/sample - loss: 0.6805 - accuracy: 0.5278 - val_loss: 0.7394 - val_accuracy: 0.4868\n",
      "Epoch 4/50\n",
      "119997/119997 [==============================] - 5s 45us/sample - loss: 0.6699 - accuracy: 0.5380 - val_loss: 0.9105 - val_accuracy: 0.5482\n",
      "Epoch 5/50\n",
      "119997/119997 [==============================] - 5s 41us/sample - loss: 0.6783 - accuracy: 0.5248 - val_loss: 1.2070 - val_accuracy: 0.4895\n",
      "Epoch 6/50\n",
      "119997/119997 [==============================] - 5s 38us/sample - loss: 0.6794 - accuracy: 0.5251 - val_loss: 1.0921 - val_accuracy: 0.5506\n",
      "Epoch 7/50\n",
      "119997/119997 [==============================] - 5s 38us/sample - loss: 0.6729 - accuracy: 0.5370 - val_loss: 0.9601 - val_accuracy: 0.4951\n",
      "Epoch 8/50\n",
      "119997/119997 [==============================] - 5s 38us/sample - loss: 0.6749 - accuracy: 0.5318 - val_loss: 0.9948 - val_accuracy: 0.5546\n",
      "Epoch 9/50\n",
      "119997/119997 [==============================] - 5s 40us/sample - loss: 0.6619 - accuracy: 0.5470 - val_loss: 1.2606 - val_accuracy: 0.4974\n",
      "Epoch 10/50\n",
      "119997/119997 [==============================] - 5s 41us/sample - loss: 0.6738 - accuracy: 0.5280 - val_loss: 1.1889 - val_accuracy: 0.5441\n",
      "Epoch 11/50\n",
      "119997/119997 [==============================] - 5s 45us/sample - loss: 0.6685 - accuracy: 0.5365 - val_loss: 1.0634 - val_accuracy: 0.5445\n",
      "Epoch 12/50\n",
      "119997/119997 [==============================] - 5s 38us/sample - loss: 0.6592 - accuracy: 0.5469 - val_loss: 1.3968 - val_accuracy: 0.5492\n",
      "Epoch 13/50\n",
      "119997/119997 [==============================] - 5s 43us/sample - loss: 0.6620 - accuracy: 0.5401 - val_loss: 1.8018 - val_accuracy: 0.6092\n",
      "Epoch 14/50\n",
      "119997/119997 [==============================] - 5s 40us/sample - loss: 0.6697 - accuracy: 0.5323 - val_loss: 1.7814 - val_accuracy: 0.5547\n",
      "Epoch 15/50\n",
      "119997/119997 [==============================] - 5s 44us/sample - loss: 0.6643 - accuracy: 0.5387 - val_loss: 1.5627 - val_accuracy: 0.5765\n",
      "Epoch 16/50\n",
      "119997/119997 [==============================] - 6s 50us/sample - loss: 0.6571 - accuracy: 0.5462 - val_loss: 1.7702 - val_accuracy: 0.5499\n",
      "Epoch 17/50\n",
      "119997/119997 [==============================] - 5s 40us/sample - loss: 0.6616 - accuracy: 0.5411 - val_loss: 1.7088 - val_accuracy: 0.5736\n",
      "Epoch 18/50\n",
      "119997/119997 [==============================] - 5s 41us/sample - loss: 0.6425 - accuracy: 0.5643 - val_loss: 2.2345 - val_accuracy: 0.5559\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 19/50\n",
      "119997/119997 [==============================] - 5s 42us/sample - loss: 0.6424 - accuracy: 0.5665 - val_loss: 2.3047 - val_accuracy: 0.5576\n",
      "Epoch 20/50\n",
      "119997/119997 [==============================] - 5s 41us/sample - loss: 0.6366 - accuracy: 0.5687 - val_loss: 1.7129 - val_accuracy: 0.5614\n",
      "Epoch 21/50\n",
      "119997/119997 [==============================] - 5s 39us/sample - loss: 0.6336 - accuracy: 0.5768 - val_loss: 2.3684 - val_accuracy: 0.5581\n",
      "Epoch 22/50\n",
      "119997/119997 [==============================] - 6s 46us/sample - loss: 0.6323 - accuracy: 0.5789 - val_loss: 1.9169 - val_accuracy: 0.5596\n",
      "Epoch 23/50\n",
      "119997/119997 [==============================] - 5s 39us/sample - loss: 0.6339 - accuracy: 0.5730 - val_loss: 2.0865 - val_accuracy: 0.5454\n",
      "Epoch 24/50\n",
      "119997/119997 [==============================] - 5s 38us/sample - loss: 0.6320 - accuracy: 0.5870 - val_loss: 2.3938 - val_accuracy: 0.5588\n",
      "Epoch 25/50\n",
      "119997/119997 [==============================] - 5s 38us/sample - loss: 0.6330 - accuracy: 0.5803 - val_loss: 1.5393 - val_accuracy: 0.5715\n",
      "Epoch 26/50\n",
      "119997/119997 [==============================] - 5s 38us/sample - loss: 0.6213 - accuracy: 0.5961 - val_loss: 2.9128 - val_accuracy: 0.5615\n",
      "Epoch 27/50\n",
      "119997/119997 [==============================] - 4s 37us/sample - loss: 0.6015 - accuracy: 0.6192 - val_loss: 3.4894 - val_accuracy: 0.5750\n",
      "Epoch 28/50\n",
      "119997/119997 [==============================] - 5s 38us/sample - loss: 0.5360 - accuracy: 0.6950 - val_loss: 4.0507 - val_accuracy: 0.5525\n",
      "Epoch 29/50\n",
      "119997/119997 [==============================] - 5s 44us/sample - loss: 0.5485 - accuracy: 0.6721 - val_loss: 5.4018 - val_accuracy: 0.5695\n",
      "Epoch 30/50\n",
      "119997/119997 [==============================] - 4s 37us/sample - loss: 0.5457 - accuracy: 0.6880 - val_loss: 3.5925 - val_accuracy: 0.5960\n",
      "Epoch 31/50\n",
      "119997/119997 [==============================] - 4s 37us/sample - loss: 0.5423 - accuracy: 0.7078 - val_loss: 6.2285 - val_accuracy: 0.5532\n",
      "Epoch 32/50\n",
      "119997/119997 [==============================] - 4s 37us/sample - loss: 0.5219 - accuracy: 0.7199 - val_loss: 7.1699 - val_accuracy: 0.5585\n",
      "Epoch 33/50\n",
      "119997/119997 [==============================] - 4s 37us/sample - loss: 0.4989 - accuracy: 0.7297 - val_loss: 8.8903 - val_accuracy: 0.5613\n",
      "Epoch 34/50\n",
      "119997/119997 [==============================] - 5s 38us/sample - loss: 0.4831 - accuracy: 0.7431 - val_loss: 12.7144 - val_accuracy: 0.5680\n",
      "Epoch 35/50\n",
      "119997/119997 [==============================] - 5s 46us/sample - loss: 0.4785 - accuracy: 0.7482 - val_loss: 18.7655 - val_accuracy: 0.5354\n",
      "Epoch 36/50\n",
      "119997/119997 [==============================] - 5s 38us/sample - loss: 0.4761 - accuracy: 0.7490 - val_loss: 17.4131 - val_accuracy: 0.5256\n",
      "Epoch 37/50\n",
      "119997/119997 [==============================] - 5s 38us/sample - loss: 0.4751 - accuracy: 0.7476 - val_loss: 20.8268 - val_accuracy: 0.5293\n",
      "Epoch 38/50\n",
      "119997/119997 [==============================] - 5s 40us/sample - loss: 0.4774 - accuracy: 0.7463 - val_loss: 22.7311 - val_accuracy: 0.5670\n",
      "Epoch 39/50\n",
      "119997/119997 [==============================] - 5s 40us/sample - loss: 0.4824 - accuracy: 0.7413 - val_loss: 23.4890 - val_accuracy: 0.5673\n",
      "Epoch 40/50\n",
      "119997/119997 [==============================] - 5s 41us/sample - loss: 0.4995 - accuracy: 0.7421 - val_loss: 14.8459 - val_accuracy: 0.5367\n",
      "Epoch 41/50\n",
      "119997/119997 [==============================] - 6s 48us/sample - loss: 0.4578 - accuracy: 0.7646 - val_loss: 21.4769 - val_accuracy: 0.5549\n",
      "Epoch 42/50\n",
      "119997/119997 [==============================] - 5s 42us/sample - loss: 0.4664 - accuracy: 0.7551 - val_loss: 22.8268 - val_accuracy: 0.5171\n",
      "Epoch 43/50\n",
      "119997/119997 [==============================] - 5s 42us/sample - loss: 0.4976 - accuracy: 0.7324 - val_loss: 22.6998 - val_accuracy: 0.5856\n",
      "Epoch 44/50\n",
      "119997/119997 [==============================] - 5s 44us/sample - loss: 0.5039 - accuracy: 0.7305 - val_loss: 9.5875 - val_accuracy: 0.5797\n",
      "Epoch 45/50\n",
      "119997/119997 [==============================] - 5s 40us/sample - loss: 0.4762 - accuracy: 0.7554 - val_loss: 13.8035 - val_accuracy: 0.6103\n",
      "Epoch 46/50\n",
      "119997/119997 [==============================] - 4s 37us/sample - loss: 0.4602 - accuracy: 0.7622 - val_loss: 17.4188 - val_accuracy: 0.5351\n",
      "Epoch 47/50\n",
      "119997/119997 [==============================] - 5s 45us/sample - loss: 0.4760 - accuracy: 0.7438 - val_loss: 17.1596 - val_accuracy: 0.5316\n",
      "Epoch 48/50\n",
      "119997/119997 [==============================] - 5s 38us/sample - loss: 0.4689 - accuracy: 0.7499 - val_loss: 25.2638 - val_accuracy: 0.5252\n",
      "Epoch 49/50\n",
      "119997/119997 [==============================] - 5s 38us/sample - loss: 0.4744 - accuracy: 0.7519 - val_loss: 16.8158 - val_accuracy: 0.5277\n",
      "Epoch 50/50\n",
      "119997/119997 [==============================] - 4s 37us/sample - loss: 0.4860 - accuracy: 0.7413 - val_loss: 18.3777 - val_accuracy: 0.5444\n",
      "3999/3999 [==============================] - 0s 21us/sample - loss: 52.8316 - accuracy: 0.4571\n",
      "3999/3999 [==============================] - 0s 21us/sample - loss: 11.1909 - accuracy: 0.6864\n",
      "3999/3999 [==============================] - 0s 22us/sample - loss: 2.8819 - accuracy: 0.8707\n",
      "5-layers-64-nodes\n",
      "Train on 119997 samples, validate on 29997 samples\n",
      "Epoch 1/50\n",
      "119997/119997 [==============================] - 6s 51us/sample - loss: 0.6904 - accuracy: 0.5069 - val_loss: 0.6578 - val_accuracy: 0.5710\n",
      "Epoch 2/50\n",
      "119997/119997 [==============================] - 5s 43us/sample - loss: 0.6856 - accuracy: 0.5139 - val_loss: 0.7132 - val_accuracy: 0.5259\n",
      "Epoch 3/50\n",
      "119997/119997 [==============================] - 6s 52us/sample - loss: 0.6772 - accuracy: 0.5272 - val_loss: 0.8211 - val_accuracy: 0.4934\n",
      "Epoch 4/50\n",
      "119997/119997 [==============================] - 5s 42us/sample - loss: 0.6708 - accuracy: 0.5353 - val_loss: 0.8705 - val_accuracy: 0.5521\n",
      "Epoch 5/50\n",
      "119997/119997 [==============================] - 5s 41us/sample - loss: 0.6683 - accuracy: 0.5373 - val_loss: 0.9351 - val_accuracy: 0.5342\n",
      "Epoch 6/50\n",
      "119997/119997 [==============================] - 5s 40us/sample - loss: 0.6673 - accuracy: 0.5386 - val_loss: 0.8659 - val_accuracy: 0.4882\n",
      "Epoch 7/50\n",
      "119997/119997 [==============================] - 5s 40us/sample - loss: 0.6647 - accuracy: 0.5414 - val_loss: 0.9663 - val_accuracy: 0.5413\n",
      "Epoch 8/50\n",
      "119997/119997 [==============================] - 5s 40us/sample - loss: 0.6637 - accuracy: 0.5449 - val_loss: 0.8567 - val_accuracy: 0.4798\n",
      "Epoch 9/50\n",
      "119997/119997 [==============================] - 6s 49us/sample - loss: 0.6656 - accuracy: 0.5364 - val_loss: 0.9441 - val_accuracy: 0.4682\n",
      "Epoch 10/50\n",
      "119997/119997 [==============================] - 5s 44us/sample - loss: 0.6689 - accuracy: 0.5385 - val_loss: 0.9575 - val_accuracy: 0.5528\n",
      "Epoch 11/50\n",
      "119997/119997 [==============================] - 5s 44us/sample - loss: 0.6004 - accuracy: 0.6329 - val_loss: 3.0756 - val_accuracy: 0.5566\n",
      "Epoch 12/50\n",
      "119997/119997 [==============================] - 5s 43us/sample - loss: 0.5351 - accuracy: 0.6954 - val_loss: 5.0649 - val_accuracy: 0.5485\n",
      "Epoch 13/50\n",
      "119997/119997 [==============================] - 5s 41us/sample - loss: 0.5194 - accuracy: 0.7097 - val_loss: 6.0616 - val_accuracy: 0.5573\n",
      "Epoch 14/50\n",
      "119997/119997 [==============================] - 5s 44us/sample - loss: 0.5166 - accuracy: 0.7122 - val_loss: 5.5367 - val_accuracy: 0.5405\n",
      "Epoch 15/50\n",
      "119997/119997 [==============================] - 5s 42us/sample - loss: 0.5121 - accuracy: 0.7152 - val_loss: 6.6948 - val_accuracy: 0.5365\n",
      "Epoch 16/50\n",
      "119997/119997 [==============================] - 5s 42us/sample - loss: 0.5041 - accuracy: 0.7215 - val_loss: 7.1025 - val_accuracy: 0.5441\n",
      "Epoch 17/50\n",
      "119997/119997 [==============================] - 5s 40us/sample - loss: 0.5090 - accuracy: 0.7219 - val_loss: 6.4855 - val_accuracy: 0.5391\n",
      "Epoch 18/50\n",
      "119997/119997 [==============================] - 5s 40us/sample - loss: 0.4998 - accuracy: 0.7274 - val_loss: 6.3669 - val_accuracy: 0.5331\n",
      "Epoch 19/50\n",
      "119997/119997 [==============================] - 5s 42us/sample - loss: 0.4939 - accuracy: 0.7316 - val_loss: 8.1489 - val_accuracy: 0.5343\n",
      "Epoch 20/50\n",
      "119997/119997 [==============================] - 5s 44us/sample - loss: 0.4993 - accuracy: 0.7255 - val_loss: 8.5572 - val_accuracy: 0.5427\n",
      "Epoch 21/50\n",
      "119997/119997 [==============================] - 5s 41us/sample - loss: 0.4911 - accuracy: 0.7362 - val_loss: 7.8514 - val_accuracy: 0.5400\n",
      "Epoch 22/50\n",
      "119997/119997 [==============================] - 5s 43us/sample - loss: 0.4973 - accuracy: 0.7317 - val_loss: 6.3818 - val_accuracy: 0.5420\n",
      "Epoch 23/50\n",
      "119997/119997 [==============================] - 6s 47us/sample - loss: 0.5003 - accuracy: 0.7273 - val_loss: 6.2729 - val_accuracy: 0.5433\n",
      "Epoch 24/50\n",
      "119997/119997 [==============================] - 5s 40us/sample - loss: 0.4923 - accuracy: 0.7358 - val_loss: 7.4887 - val_accuracy: 0.5488\n",
      "Epoch 25/50\n",
      "119997/119997 [==============================] - 5s 40us/sample - loss: 0.4844 - accuracy: 0.7433 - val_loss: 7.0655 - val_accuracy: 0.5419\n",
      "Epoch 26/50\n",
      "119997/119997 [==============================] - 5s 44us/sample - loss: 0.4710 - accuracy: 0.7522 - val_loss: 8.4605 - val_accuracy: 0.5420\n",
      "Epoch 27/50\n",
      "119997/119997 [==============================] - 5s 41us/sample - loss: 0.4942 - accuracy: 0.7333 - val_loss: 7.7829 - val_accuracy: 0.5294\n",
      "Epoch 28/50\n",
      "119997/119997 [==============================] - 5s 40us/sample - loss: 0.4861 - accuracy: 0.7423 - val_loss: 10.0194 - val_accuracy: 0.5284\n",
      "Epoch 29/50\n",
      "119997/119997 [==============================] - 5s 40us/sample - loss: 0.4911 - accuracy: 0.7447 - val_loss: 8.0156 - val_accuracy: 0.5150\n",
      "Epoch 30/50\n",
      "119997/119997 [==============================] - 5s 40us/sample - loss: 0.4738 - accuracy: 0.7589 - val_loss: 12.6647 - val_accuracy: 0.5220\n",
      "Epoch 31/50\n",
      "119997/119997 [==============================] - 5s 41us/sample - loss: 0.4550 - accuracy: 0.7725 - val_loss: 12.3199 - val_accuracy: 0.5515\n",
      "Epoch 32/50\n",
      "119997/119997 [==============================] - 5s 44us/sample - loss: 0.4564 - accuracy: 0.7716 - val_loss: 16.1527 - val_accuracy: 0.5178\n",
      "Epoch 33/50\n",
      "119997/119997 [==============================] - 5s 44us/sample - loss: 0.4635 - accuracy: 0.7620 - val_loss: 20.2899 - val_accuracy: 0.5273\n",
      "Epoch 34/50\n",
      "119997/119997 [==============================] - 5s 44us/sample - loss: 0.4529 - accuracy: 0.7733 - val_loss: 17.5729 - val_accuracy: 0.5261\n",
      "Epoch 35/50\n",
      "119997/119997 [==============================] - 5s 45us/sample - loss: 0.4505 - accuracy: 0.7740 - val_loss: 20.1057 - val_accuracy: 0.5273\n",
      "Epoch 36/50\n",
      "119997/119997 [==============================] - 5s 40us/sample - loss: 0.4468 - accuracy: 0.7781 - val_loss: 29.3275 - val_accuracy: 0.5539\n",
      "Epoch 37/50\n",
      "119997/119997 [==============================] - 5s 40us/sample - loss: 0.4330 - accuracy: 0.7869 - val_loss: 24.9017 - val_accuracy: 0.5551\n",
      "Epoch 38/50\n",
      "119997/119997 [==============================] - 6s 46us/sample - loss: 0.4437 - accuracy: 0.7783 - val_loss: 23.1722 - val_accuracy: 0.5245\n",
      "Epoch 39/50\n",
      "119997/119997 [==============================] - 5s 42us/sample - loss: 0.4360 - accuracy: 0.7843 - val_loss: 32.4886 - val_accuracy: 0.5445\n",
      "Epoch 40/50\n",
      "119997/119997 [==============================] - 5s 40us/sample - loss: 0.4434 - accuracy: 0.7786 - val_loss: 22.5962 - val_accuracy: 0.5541\n",
      "Epoch 41/50\n",
      "119997/119997 [==============================] - 5s 40us/sample - loss: 0.4306 - accuracy: 0.7858 - val_loss: 9.5266 - val_accuracy: 0.5522\n",
      "Epoch 42/50\n",
      "119997/119997 [==============================] - 5s 40us/sample - loss: 0.4543 - accuracy: 0.7647 - val_loss: 6.6152 - val_accuracy: 0.5586\n",
      "Epoch 43/50\n",
      "119997/119997 [==============================] - 5s 40us/sample - loss: 0.4735 - accuracy: 0.7492 - val_loss: 10.6109 - val_accuracy: 0.5513\n",
      "Epoch 44/50\n",
      "119997/119997 [==============================] - 5s 45us/sample - loss: 0.4636 - accuracy: 0.7562 - val_loss: 14.8115 - val_accuracy: 0.5491\n",
      "Epoch 45/50\n",
      "119997/119997 [==============================] - 5s 42us/sample - loss: 0.4378 - accuracy: 0.7798 - val_loss: 16.5553 - val_accuracy: 0.5410\n",
      "Epoch 46/50\n",
      "119997/119997 [==============================] - 5s 40us/sample - loss: 0.4481 - accuracy: 0.7715 - val_loss: 15.7183 - val_accuracy: 0.5199\n",
      "Epoch 47/50\n",
      "119997/119997 [==============================] - 5s 41us/sample - loss: 0.4407 - accuracy: 0.7799 - val_loss: 15.0923 - val_accuracy: 0.5372\n",
      "Epoch 48/50\n",
      "119997/119997 [==============================] - 5s 41us/sample - loss: 0.4842 - accuracy: 0.7353 - val_loss: 13.7741 - val_accuracy: 0.5451\n",
      "Epoch 49/50\n",
      "119997/119997 [==============================] - 5s 46us/sample - loss: 0.5004 - accuracy: 0.7245 - val_loss: 9.4065 - val_accuracy: 0.5260\n",
      "Epoch 50/50\n",
      "119997/119997 [==============================] - 6s 51us/sample - loss: 0.4838 - accuracy: 0.7389 - val_loss: 16.2951 - val_accuracy: 0.5581\n",
      "3999/3999 [==============================] - 0s 23us/sample - loss: 56.9686 - accuracy: 0.4556\n",
      "3999/3999 [==============================] - 0s 21us/sample - loss: 2.9484 - accuracy: 0.7827\n",
      "3999/3999 [==============================] - 0s 21us/sample - loss: 0.4895 - accuracy: 0.9272\n",
      "5-layers-128-nodes\n",
      "Train on 119997 samples, validate on 29997 samples\n",
      "Epoch 1/50\n",
      "119997/119997 [==============================] - 7s 62us/sample - loss: 0.6909 - accuracy: 0.5055 - val_loss: 0.6710 - val_accuracy: 0.5272\n",
      "Epoch 2/50\n",
      "119997/119997 [==============================] - 7s 54us/sample - loss: 0.6901 - accuracy: 0.5035 - val_loss: 0.6732 - val_accuracy: 0.5318\n",
      "Epoch 3/50\n",
      "119997/119997 [==============================] - 7s 54us/sample - loss: 0.6899 - accuracy: 0.5029 - val_loss: 0.6690 - val_accuracy: 0.4999\n",
      "Epoch 4/50\n",
      "119997/119997 [==============================] - 7s 56us/sample - loss: 0.6898 - accuracy: 0.5039 - val_loss: 0.6727 - val_accuracy: 0.5303\n",
      "Epoch 5/50\n",
      "119997/119997 [==============================] - 7s 55us/sample - loss: 0.6899 - accuracy: 0.5048 - val_loss: 0.6721 - val_accuracy: 0.4999\n",
      "Epoch 6/50\n",
      "119997/119997 [==============================] - 7s 55us/sample - loss: 0.6894 - accuracy: 0.5051 - val_loss: 0.6766 - val_accuracy: 0.5245\n",
      "Epoch 7/50\n",
      "119997/119997 [==============================] - 7s 56us/sample - loss: 0.6898 - accuracy: 0.5059 - val_loss: 0.6779 - val_accuracy: 0.5222\n",
      "Epoch 8/50\n",
      "119997/119997 [==============================] - 6s 54us/sample - loss: 0.6896 - accuracy: 0.5042 - val_loss: 0.6741 - val_accuracy: 0.5276\n",
      "Epoch 9/50\n",
      "119997/119997 [==============================] - 7s 59us/sample - loss: 0.6899 - accuracy: 0.5042 - val_loss: 0.6697 - val_accuracy: 0.5354\n",
      "Epoch 10/50\n",
      "119997/119997 [==============================] - 7s 59us/sample - loss: 0.6895 - accuracy: 0.5041 - val_loss: 0.6715 - val_accuracy: 0.5317\n",
      "Epoch 11/50\n",
      "119997/119997 [==============================] - 6s 54us/sample - loss: 0.6893 - accuracy: 0.5050 - val_loss: 0.6754 - val_accuracy: 0.5342\n",
      "Epoch 12/50\n",
      "119997/119997 [==============================] - 6s 47us/sample - loss: 0.6903 - accuracy: 0.5036 - val_loss: 0.6705 - val_accuracy: 0.5333\n",
      "Epoch 13/50\n",
      "119997/119997 [==============================] - 6s 48us/sample - loss: 0.6907 - accuracy: 0.5044 - val_loss: 0.6726 - val_accuracy: 0.5322\n",
      "Epoch 14/50\n",
      "119997/119997 [==============================] - 7s 55us/sample - loss: 0.6896 - accuracy: 0.5061 - val_loss: 0.6710 - val_accuracy: 0.4999\n",
      "Epoch 15/50\n",
      "119997/119997 [==============================] - 6s 53us/sample - loss: 0.6892 - accuracy: 0.5042 - val_loss: 0.6731 - val_accuracy: 0.5291\n",
      "Epoch 16/50\n",
      "119997/119997 [==============================] - 7s 55us/sample - loss: 0.6894 - accuracy: 0.5040 - val_loss: 0.6694 - val_accuracy: 0.5395\n",
      "Epoch 17/50\n",
      "119997/119997 [==============================] - 6s 52us/sample - loss: 0.6897 - accuracy: 0.5064 - val_loss: 0.6715 - val_accuracy: 0.4999\n",
      "Epoch 18/50\n",
      "119997/119997 [==============================] - 7s 59us/sample - loss: 0.6892 - accuracy: 0.5029 - val_loss: 0.6722 - val_accuracy: 0.5306\n",
      "Epoch 19/50\n",
      "119997/119997 [==============================] - 6s 53us/sample - loss: 0.6897 - accuracy: 0.5046 - val_loss: 0.6693 - val_accuracy: 0.5363\n",
      "Epoch 20/50\n",
      "119997/119997 [==============================] - 6s 50us/sample - loss: 0.6891 - accuracy: 0.5048 - val_loss: 0.6709 - val_accuracy: 0.4997\n",
      "Epoch 21/50\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "119997/119997 [==============================] - 6s 54us/sample - loss: 0.6893 - accuracy: 0.5049 - val_loss: 0.6716 - val_accuracy: 0.5339\n",
      "Epoch 22/50\n",
      "119997/119997 [==============================] - 6s 50us/sample - loss: 0.6892 - accuracy: 0.5045 - val_loss: 0.6707 - val_accuracy: 0.5563\n",
      "Epoch 23/50\n",
      "119997/119997 [==============================] - 7s 54us/sample - loss: 0.6894 - accuracy: 0.5060 - val_loss: 0.6672 - val_accuracy: 0.5414\n",
      "Epoch 24/50\n",
      "119997/119997 [==============================] - 6s 48us/sample - loss: 0.6893 - accuracy: 0.5050 - val_loss: 0.6687 - val_accuracy: 0.5373\n",
      "Epoch 25/50\n",
      "119997/119997 [==============================] - 7s 57us/sample - loss: 0.6918 - accuracy: 0.5046 - val_loss: 0.6743 - val_accuracy: 0.5287\n",
      "Epoch 26/50\n",
      "119997/119997 [==============================] - 6s 53us/sample - loss: 0.6892 - accuracy: 0.5055 - val_loss: 0.6700 - val_accuracy: 0.5380\n",
      "Epoch 27/50\n",
      "119997/119997 [==============================] - 6s 51us/sample - loss: 0.6892 - accuracy: 0.5050 - val_loss: 0.6698 - val_accuracy: 0.4999\n",
      "Epoch 28/50\n",
      "119997/119997 [==============================] - 7s 58us/sample - loss: 0.6892 - accuracy: 0.5037 - val_loss: 0.6726 - val_accuracy: 0.5371\n",
      "Epoch 29/50\n",
      "119997/119997 [==============================] - 6s 54us/sample - loss: 0.6895 - accuracy: 0.5038 - val_loss: 0.6734 - val_accuracy: 0.4996\n",
      "Epoch 30/50\n",
      "119997/119997 [==============================] - 6s 48us/sample - loss: 0.6892 - accuracy: 0.5051 - val_loss: 0.6723 - val_accuracy: 0.5321\n",
      "Epoch 31/50\n",
      "119997/119997 [==============================] - 6s 48us/sample - loss: 0.6891 - accuracy: 0.5050 - val_loss: 0.6704 - val_accuracy: 0.5391\n",
      "Epoch 32/50\n",
      "119997/119997 [==============================] - 6s 50us/sample - loss: 0.6889 - accuracy: 0.5061 - val_loss: 0.6708 - val_accuracy: 0.5342\n",
      "Epoch 33/50\n",
      "119997/119997 [==============================] - 6s 50us/sample - loss: 0.6891 - accuracy: 0.5040 - val_loss: 0.6701 - val_accuracy: 0.5377\n",
      "Epoch 34/50\n",
      "119997/119997 [==============================] - 6s 51us/sample - loss: 0.6894 - accuracy: 0.5060 - val_loss: 0.6692 - val_accuracy: 0.5406\n",
      "Epoch 35/50\n",
      "119997/119997 [==============================] - 6s 47us/sample - loss: 0.6886 - accuracy: 0.5046 - val_loss: 0.6695 - val_accuracy: 0.5374\n",
      "Epoch 36/50\n",
      "119997/119997 [==============================] - 6s 51us/sample - loss: 0.6891 - accuracy: 0.5031 - val_loss: 0.6703 - val_accuracy: 0.5370\n",
      "Epoch 37/50\n",
      "119997/119997 [==============================] - 7s 57us/sample - loss: 0.6892 - accuracy: 0.5049 - val_loss: 0.6700 - val_accuracy: 0.5397\n",
      "Epoch 38/50\n",
      "119997/119997 [==============================] - 6s 51us/sample - loss: 0.6888 - accuracy: 0.5061 - val_loss: 0.6691 - val_accuracy: 0.5408\n",
      "Epoch 39/50\n",
      "119997/119997 [==============================] - 7s 56us/sample - loss: 0.6888 - accuracy: 0.5059 - val_loss: 0.6687 - val_accuracy: 0.4999\n",
      "Epoch 40/50\n",
      "119997/119997 [==============================] - 7s 56us/sample - loss: 0.6889 - accuracy: 0.5055 - val_loss: 0.6709 - val_accuracy: 0.4999\n",
      "Epoch 41/50\n",
      "119997/119997 [==============================] - 6s 52us/sample - loss: 0.6889 - accuracy: 0.5046 - val_loss: 0.6708 - val_accuracy: 0.4999\n",
      "Epoch 42/50\n",
      "119997/119997 [==============================] - 6s 53us/sample - loss: 0.6888 - accuracy: 0.5056 - val_loss: 0.6703 - val_accuracy: 0.5130\n",
      "Epoch 43/50\n",
      "119997/119997 [==============================] - 6s 49us/sample - loss: 0.6885 - accuracy: 0.5066 - val_loss: 0.6680 - val_accuracy: 0.5448\n",
      "Epoch 44/50\n",
      "119997/119997 [==============================] - 6s 52us/sample - loss: 0.6887 - accuracy: 0.5058 - val_loss: 0.6696 - val_accuracy: 0.5399\n",
      "Epoch 45/50\n",
      "119997/119997 [==============================] - 6s 51us/sample - loss: 0.6890 - accuracy: 0.5051 - val_loss: 0.6733 - val_accuracy: 0.5463\n",
      "Epoch 46/50\n",
      "119997/119997 [==============================] - 6s 51us/sample - loss: 0.6893 - accuracy: 0.5042 - val_loss: 0.6703 - val_accuracy: 0.5394\n",
      "Epoch 47/50\n",
      "119997/119997 [==============================] - 7s 58us/sample - loss: 0.6879 - accuracy: 0.5087 - val_loss: 0.6644 - val_accuracy: 0.5523\n",
      "Epoch 48/50\n",
      "119997/119997 [==============================] - 6s 52us/sample - loss: 0.6885 - accuracy: 0.5058 - val_loss: 0.6728 - val_accuracy: 0.5307\n",
      "Epoch 49/50\n",
      "119997/119997 [==============================] - 6s 52us/sample - loss: 0.6892 - accuracy: 0.5059 - val_loss: 0.6724 - val_accuracy: 0.5314\n",
      "Epoch 50/50\n",
      "119997/119997 [==============================] - 6s 48us/sample - loss: 0.6887 - accuracy: 0.5062 - val_loss: 0.6649 - val_accuracy: 0.5470\n",
      "3999/3999 [==============================] - 0s 21us/sample - loss: 0.7851 - accuracy: 0.4576\n",
      "3999/3999 [==============================] - 0s 21us/sample - loss: 0.6353 - accuracy: 0.5876\n",
      "3999/3999 [==============================] - 0s 20us/sample - loss: 0.5823 - accuracy: 0.6684\n",
      "6-layers-16-nodes\n",
      "Train on 119997 samples, validate on 29997 samples\n",
      "Epoch 1/50\n",
      "119997/119997 [==============================] - 6s 47us/sample - loss: 0.6910 - accuracy: 0.5007 - val_loss: 0.6665 - val_accuracy: 0.5491\n",
      "Epoch 2/50\n",
      "119997/119997 [==============================] - 6s 50us/sample - loss: 0.6897 - accuracy: 0.5023 - val_loss: 0.8011 - val_accuracy: 0.5000\n",
      "Epoch 3/50\n",
      "119997/119997 [==============================] - 5s 45us/sample - loss: 0.6895 - accuracy: 0.5049 - val_loss: 0.6535 - val_accuracy: 0.5635\n",
      "Epoch 4/50\n",
      "119997/119997 [==============================] - 5s 41us/sample - loss: 0.6895 - accuracy: 0.5051 - val_loss: 0.6553 - val_accuracy: 0.4999\n",
      "Epoch 5/50\n",
      "119997/119997 [==============================] - 5s 42us/sample - loss: 0.6895 - accuracy: 0.5033 - val_loss: 0.9768 - val_accuracy: 0.5035\n",
      "Epoch 6/50\n",
      "119997/119997 [==============================] - 5s 42us/sample - loss: 0.6895 - accuracy: 0.5056 - val_loss: 0.6602 - val_accuracy: 0.5532\n",
      "Epoch 7/50\n",
      "119997/119997 [==============================] - 5s 43us/sample - loss: 0.6893 - accuracy: 0.5047 - val_loss: 0.6538 - val_accuracy: 0.5461\n",
      "Epoch 8/50\n",
      "119997/119997 [==============================] - 5s 39us/sample - loss: 0.6895 - accuracy: 0.5045 - val_loss: 0.6665 - val_accuracy: 0.5403\n",
      "Epoch 9/50\n",
      "119997/119997 [==============================] - 5s 41us/sample - loss: 0.6895 - accuracy: 0.5034 - val_loss: 0.6670 - val_accuracy: 0.5385\n",
      "Epoch 10/50\n",
      "119997/119997 [==============================] - 5s 43us/sample - loss: 0.6894 - accuracy: 0.5055 - val_loss: 0.6675 - val_accuracy: 0.5384\n",
      "Epoch 11/50\n",
      "119997/119997 [==============================] - 5s 42us/sample - loss: 0.6893 - accuracy: 0.5068 - val_loss: 0.6494 - val_accuracy: 0.5637\n",
      "Epoch 12/50\n",
      "119997/119997 [==============================] - 5s 42us/sample - loss: 0.6889 - accuracy: 0.5084 - val_loss: 0.6512 - val_accuracy: 0.5561\n",
      "Epoch 13/50\n",
      "119997/119997 [==============================] - 5s 46us/sample - loss: 0.6890 - accuracy: 0.5085 - val_loss: 0.6590 - val_accuracy: 0.5525\n",
      "Epoch 14/50\n",
      "119997/119997 [==============================] - 5s 41us/sample - loss: 0.6892 - accuracy: 0.5073 - val_loss: 0.6660 - val_accuracy: 0.5313\n",
      "Epoch 15/50\n",
      "119997/119997 [==============================] - 5s 42us/sample - loss: 0.6890 - accuracy: 0.5095 - val_loss: 0.6526 - val_accuracy: 0.5526\n",
      "Epoch 16/50\n",
      "119997/119997 [==============================] - 5s 44us/sample - loss: 0.6890 - accuracy: 0.5103 - val_loss: 0.6564 - val_accuracy: 0.5615\n",
      "Epoch 17/50\n",
      "119997/119997 [==============================] - 5s 39us/sample - loss: 0.6890 - accuracy: 0.5110 - val_loss: 0.6540 - val_accuracy: 0.5314\n",
      "Epoch 18/50\n",
      "119997/119997 [==============================] - 5s 39us/sample - loss: 0.6880 - accuracy: 0.5138 - val_loss: 0.6535 - val_accuracy: 0.5571\n",
      "Epoch 19/50\n",
      "119997/119997 [==============================] - 5s 43us/sample - loss: 0.6887 - accuracy: 0.5167 - val_loss: 0.6552 - val_accuracy: 0.5379\n",
      "Epoch 20/50\n",
      "119997/119997 [==============================] - 5s 39us/sample - loss: 0.6882 - accuracy: 0.5193 - val_loss: 0.6587 - val_accuracy: 0.5354\n",
      "Epoch 21/50\n",
      "119997/119997 [==============================] - 5s 39us/sample - loss: 0.6873 - accuracy: 0.5218 - val_loss: 0.6647 - val_accuracy: 0.5072\n",
      "Epoch 22/50\n",
      "119997/119997 [==============================] - 5s 39us/sample - loss: 0.6857 - accuracy: 0.5202 - val_loss: 0.6926 - val_accuracy: 0.4629\n",
      "Epoch 23/50\n",
      "119997/119997 [==============================] - 5s 41us/sample - loss: 0.6834 - accuracy: 0.5241 - val_loss: 0.6996 - val_accuracy: 0.4746\n",
      "Epoch 24/50\n",
      "119997/119997 [==============================] - 5s 40us/sample - loss: 0.6727 - accuracy: 0.5428 - val_loss: 0.7401 - val_accuracy: 0.4775\n",
      "Epoch 25/50\n",
      "119997/119997 [==============================] - 5s 43us/sample - loss: 0.6247 - accuracy: 0.6116 - val_loss: 0.9900 - val_accuracy: 0.5395\n",
      "Epoch 26/50\n",
      "119997/119997 [==============================] - 5s 41us/sample - loss: 0.5945 - accuracy: 0.6542 - val_loss: 1.2771 - val_accuracy: 0.5414\n",
      "Epoch 27/50\n",
      "119997/119997 [==============================] - 5s 39us/sample - loss: 0.5889 - accuracy: 0.6535 - val_loss: 1.3247 - val_accuracy: 0.5384\n",
      "Epoch 28/50\n",
      "119997/119997 [==============================] - 5s 39us/sample - loss: 0.5865 - accuracy: 0.6542 - val_loss: 1.0125 - val_accuracy: 0.5420\n",
      "Epoch 29/50\n",
      "119997/119997 [==============================] - 5s 40us/sample - loss: 0.5885 - accuracy: 0.6523 - val_loss: 0.9106 - val_accuracy: 0.5569\n",
      "Epoch 30/50\n",
      "119997/119997 [==============================] - 5s 45us/sample - loss: 0.5866 - accuracy: 0.6545 - val_loss: 1.3287 - val_accuracy: 0.5482\n",
      "Epoch 31/50\n",
      "119997/119997 [==============================] - 6s 48us/sample - loss: 0.6657 - accuracy: 0.5683 - val_loss: 0.7925 - val_accuracy: 0.4798\n",
      "Epoch 32/50\n",
      "119997/119997 [==============================] - 6s 47us/sample - loss: 0.6807 - accuracy: 0.5477 - val_loss: 1.1011 - val_accuracy: 0.4778\n",
      "Epoch 33/50\n",
      "119997/119997 [==============================] - 5s 43us/sample - loss: 0.6660 - accuracy: 0.5621 - val_loss: 0.8075 - val_accuracy: 0.4791\n",
      "Epoch 34/50\n",
      "119997/119997 [==============================] - 6s 48us/sample - loss: 0.6184 - accuracy: 0.6193 - val_loss: 2.2188 - val_accuracy: 0.5373\n",
      "Epoch 35/50\n",
      "119997/119997 [==============================] - 6s 46us/sample - loss: 0.5864 - accuracy: 0.6574 - val_loss: 2.9150 - val_accuracy: 0.5288\n",
      "Epoch 36/50\n",
      "119997/119997 [==============================] - 6s 46us/sample - loss: 0.5764 - accuracy: 0.6687 - val_loss: 5.6849 - val_accuracy: 0.5300\n",
      "Epoch 37/50\n",
      "119997/119997 [==============================] - 6s 49us/sample - loss: 0.5638 - accuracy: 0.6806 - val_loss: 5.3771 - val_accuracy: 0.5692\n",
      "Epoch 38/50\n",
      "119997/119997 [==============================] - 5s 45us/sample - loss: 0.5595 - accuracy: 0.6830 - val_loss: 7.3576 - val_accuracy: 0.5474\n",
      "Epoch 39/50\n",
      "119997/119997 [==============================] - 6s 48us/sample - loss: 0.5567 - accuracy: 0.6862 - val_loss: 7.0893 - val_accuracy: 0.5612\n",
      "Epoch 40/50\n",
      "119997/119997 [==============================] - 5s 45us/sample - loss: 0.5584 - accuracy: 0.6871 - val_loss: 7.2897 - val_accuracy: 0.5554\n",
      "Epoch 41/50\n",
      "119997/119997 [==============================] - 5s 46us/sample - loss: 0.5560 - accuracy: 0.6897 - val_loss: 8.4566 - val_accuracy: 0.5570\n",
      "Epoch 42/50\n",
      "119997/119997 [==============================] - 6s 49us/sample - loss: 0.5560 - accuracy: 0.6831 - val_loss: 7.6247 - val_accuracy: 0.5610\n",
      "Epoch 43/50\n",
      "119997/119997 [==============================] - 5s 45us/sample - loss: 0.5518 - accuracy: 0.7135 - val_loss: 10.8568 - val_accuracy: 0.5389\n",
      "Epoch 44/50\n",
      "119997/119997 [==============================] - 6s 46us/sample - loss: 0.5436 - accuracy: 0.6985 - val_loss: 10.4211 - val_accuracy: 0.5609\n",
      "Epoch 45/50\n",
      "119997/119997 [==============================] - 5s 45us/sample - loss: 0.5474 - accuracy: 0.6931 - val_loss: 10.0568 - val_accuracy: 0.5457\n",
      "Epoch 46/50\n",
      "119997/119997 [==============================] - 5s 46us/sample - loss: 0.5453 - accuracy: 0.6952 - val_loss: 11.9125 - val_accuracy: 0.5421\n",
      "Epoch 47/50\n",
      "119997/119997 [==============================] - 5s 45us/sample - loss: 0.5272 - accuracy: 0.7254 - val_loss: 11.2485 - val_accuracy: 0.5335\n",
      "Epoch 48/50\n",
      "119997/119997 [==============================] - 6s 49us/sample - loss: 0.5240 - accuracy: 0.7208 - val_loss: 11.6369 - val_accuracy: 0.5430\n",
      "Epoch 49/50\n",
      "119997/119997 [==============================] - 5s 46us/sample - loss: 0.5131 - accuracy: 0.7286 - val_loss: 13.2469 - val_accuracy: 0.5369\n",
      "Epoch 50/50\n",
      "119997/119997 [==============================] - 6s 46us/sample - loss: 0.5191 - accuracy: 0.7226 - val_loss: 12.4903 - val_accuracy: 0.5297\n",
      "3999/3999 [==============================] - 0s 20us/sample - loss: 2.7136 - accuracy: 0.4006\n",
      "3999/3999 [==============================] - 0s 21us/sample - loss: 17.8663 - accuracy: 0.5816\n",
      "3999/3999 [==============================] - 0s 19us/sample - loss: 10.4767 - accuracy: 0.7139\n",
      "6-layers-32-nodes\n",
      "Train on 119997 samples, validate on 29997 samples\n",
      "Epoch 1/50\n",
      "119997/119997 [==============================] - 6s 52us/sample - loss: 0.6912 - accuracy: 0.5020 - val_loss: 0.6637 - val_accuracy: 0.4999\n",
      "Epoch 2/50\n",
      "119997/119997 [==============================] - 5s 45us/sample - loss: 0.6897 - accuracy: 0.5047 - val_loss: 0.6728 - val_accuracy: 0.4999\n",
      "Epoch 3/50\n",
      "119997/119997 [==============================] - 6s 49us/sample - loss: 0.6895 - accuracy: 0.5049 - val_loss: 0.7235 - val_accuracy: 0.4972\n",
      "Epoch 4/50\n",
      "119997/119997 [==============================] - 6s 47us/sample - loss: 0.6894 - accuracy: 0.5058 - val_loss: 0.6604 - val_accuracy: 0.5608\n",
      "Epoch 5/50\n",
      "119997/119997 [==============================] - 6s 53us/sample - loss: 0.6887 - accuracy: 0.5122 - val_loss: 0.7026 - val_accuracy: 0.5130\n",
      "Epoch 6/50\n",
      "119997/119997 [==============================] - 6s 50us/sample - loss: 0.6851 - accuracy: 0.5213 - val_loss: 0.7255 - val_accuracy: 0.4972\n",
      "Epoch 7/50\n",
      "119997/119997 [==============================] - 6s 49us/sample - loss: 0.6812 - accuracy: 0.5245 - val_loss: 0.7748 - val_accuracy: 0.5217\n",
      "Epoch 8/50\n",
      "119997/119997 [==============================] - 6s 53us/sample - loss: 0.6714 - accuracy: 0.5356 - val_loss: 0.9100 - val_accuracy: 0.5226\n",
      "Epoch 9/50\n",
      "119997/119997 [==============================] - 6s 47us/sample - loss: 0.6699 - accuracy: 0.5349 - val_loss: 1.0688 - val_accuracy: 0.5040\n",
      "Epoch 10/50\n",
      "119997/119997 [==============================] - 6s 49us/sample - loss: 0.6721 - accuracy: 0.5332 - val_loss: 0.9111 - val_accuracy: 0.5927\n",
      "Epoch 11/50\n",
      "119997/119997 [==============================] - 5s 45us/sample - loss: 0.6617 - accuracy: 0.5566 - val_loss: 2.0749 - val_accuracy: 0.5593\n",
      "Epoch 12/50\n",
      "119997/119997 [==============================] - 5s 45us/sample - loss: 0.5682 - accuracy: 0.6574 - val_loss: 3.5862 - val_accuracy: 0.5463\n",
      "Epoch 13/50\n",
      "119997/119997 [==============================] - 6s 48us/sample - loss: 0.5622 - accuracy: 0.6622 - val_loss: 4.9666 - val_accuracy: 0.5295\n",
      "Epoch 14/50\n",
      "119997/119997 [==============================] - 5s 45us/sample - loss: 0.5611 - accuracy: 0.6648 - val_loss: 6.4303 - val_accuracy: 0.5418\n",
      "Epoch 15/50\n",
      "119997/119997 [==============================] - 5s 45us/sample - loss: 0.5438 - accuracy: 0.6858 - val_loss: 6.6235 - val_accuracy: 0.5173\n",
      "Epoch 16/50\n",
      "119997/119997 [==============================] - 5s 44us/sample - loss: 0.5139 - accuracy: 0.7166 - val_loss: 7.6983 - val_accuracy: 0.5411\n",
      "Epoch 17/50\n",
      "119997/119997 [==============================] - 5s 45us/sample - loss: 0.5151 - accuracy: 0.7111 - val_loss: 8.7251 - val_accuracy: 0.5299\n",
      "Epoch 18/50\n",
      "119997/119997 [==============================] - 6s 47us/sample - loss: 0.5076 - accuracy: 0.7231 - val_loss: 10.0792 - val_accuracy: 0.5351\n",
      "Epoch 19/50\n",
      "119997/119997 [==============================] - 5s 45us/sample - loss: 0.5052 - accuracy: 0.7256 - val_loss: 8.2861 - val_accuracy: 0.5217\n",
      "Epoch 20/50\n",
      "119997/119997 [==============================] - 5s 45us/sample - loss: 0.5042 - accuracy: 0.7191 - val_loss: 10.0534 - val_accuracy: 0.5280\n",
      "Epoch 21/50\n",
      "119997/119997 [==============================] - 5s 45us/sample - loss: 0.5052 - accuracy: 0.7178 - val_loss: 14.1653 - val_accuracy: 0.5239\n",
      "Epoch 22/50\n",
      "119997/119997 [==============================] - 5s 45us/sample - loss: 0.5072 - accuracy: 0.7158 - val_loss: 11.7327 - val_accuracy: 0.5228\n",
      "Epoch 23/50\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "119997/119997 [==============================] - 5s 45us/sample - loss: 0.5075 - accuracy: 0.7156 - val_loss: 16.5546 - val_accuracy: 0.5137\n",
      "Epoch 24/50\n",
      "119997/119997 [==============================] - 6s 49us/sample - loss: 0.5085 - accuracy: 0.7134 - val_loss: 9.7050 - val_accuracy: 0.5184\n",
      "Epoch 25/50\n",
      "119997/119997 [==============================] - 6s 46us/sample - loss: 0.5040 - accuracy: 0.7193 - val_loss: 15.0641 - val_accuracy: 0.5216\n",
      "Epoch 26/50\n",
      "119997/119997 [==============================] - 5s 45us/sample - loss: 0.5057 - accuracy: 0.7178 - val_loss: 8.2435 - val_accuracy: 0.5269\n",
      "Epoch 27/50\n",
      "119997/119997 [==============================] - 5s 45us/sample - loss: 0.5034 - accuracy: 0.7188 - val_loss: 17.0218 - val_accuracy: 0.5217\n",
      "Epoch 28/50\n",
      "119997/119997 [==============================] - 5s 45us/sample - loss: 0.5047 - accuracy: 0.7147 - val_loss: 18.4233 - val_accuracy: 0.5117\n",
      "Epoch 29/50\n",
      "119997/119997 [==============================] - 6s 49us/sample - loss: 0.5093 - accuracy: 0.7143 - val_loss: 14.6791 - val_accuracy: 0.5244\n",
      "Epoch 30/50\n",
      "119997/119997 [==============================] - 5s 45us/sample - loss: 0.5089 - accuracy: 0.7118 - val_loss: 13.3386 - val_accuracy: 0.5187\n",
      "Epoch 31/50\n",
      "119997/119997 [==============================] - 5s 45us/sample - loss: 0.5028 - accuracy: 0.7183 - val_loss: 14.0026 - val_accuracy: 0.5202\n",
      "Epoch 32/50\n",
      "119997/119997 [==============================] - 6s 49us/sample - loss: 0.5012 - accuracy: 0.7171 - val_loss: 12.3529 - val_accuracy: 0.5282\n",
      "Epoch 33/50\n",
      "119997/119997 [==============================] - 5s 45us/sample - loss: 0.5018 - accuracy: 0.7163 - val_loss: 11.4896 - val_accuracy: 0.5223\n",
      "Epoch 34/50\n",
      "119997/119997 [==============================] - 5s 44us/sample - loss: 0.4995 - accuracy: 0.7192 - val_loss: 9.9498 - val_accuracy: 0.5294\n",
      "Epoch 35/50\n",
      "119997/119997 [==============================] - 5s 44us/sample - loss: 0.5007 - accuracy: 0.7192 - val_loss: 14.5839 - val_accuracy: 0.5160\n",
      "Epoch 36/50\n",
      "119997/119997 [==============================] - 5s 39us/sample - loss: 0.5009 - accuracy: 0.7168 - val_loss: 10.7209 - val_accuracy: 0.5299\n",
      "Epoch 37/50\n",
      "119997/119997 [==============================] - 5s 39us/sample - loss: 0.4970 - accuracy: 0.7214 - val_loss: 11.6532 - val_accuracy: 0.5204\n",
      "Epoch 38/50\n",
      "119997/119997 [==============================] - 5s 40us/sample - loss: 0.4987 - accuracy: 0.7185 - val_loss: 13.5548 - val_accuracy: 0.5197\n",
      "Epoch 39/50\n",
      "119997/119997 [==============================] - 5s 42us/sample - loss: 0.5019 - accuracy: 0.7175 - val_loss: 14.2042 - val_accuracy: 0.5285\n",
      "Epoch 40/50\n",
      "119997/119997 [==============================] - 5s 42us/sample - loss: 0.5017 - accuracy: 0.7217 - val_loss: 13.9730 - val_accuracy: 0.5233\n",
      "Epoch 41/50\n",
      "119997/119997 [==============================] - 5s 45us/sample - loss: 0.5003 - accuracy: 0.7211 - val_loss: 13.0333 - val_accuracy: 0.5167\n",
      "Epoch 42/50\n",
      "119997/119997 [==============================] - 5s 40us/sample - loss: 0.4869 - accuracy: 0.7328 - val_loss: 12.7087 - val_accuracy: 0.5223\n",
      "Epoch 43/50\n",
      "119997/119997 [==============================] - 5s 40us/sample - loss: 0.4956 - accuracy: 0.7227 - val_loss: 20.1407 - val_accuracy: 0.5112\n",
      "Epoch 44/50\n",
      "119997/119997 [==============================] - 5s 40us/sample - loss: 0.4903 - accuracy: 0.7276 - val_loss: 15.9635 - val_accuracy: 0.5187\n",
      "Epoch 45/50\n",
      "119997/119997 [==============================] - 5s 39us/sample - loss: 0.4997 - accuracy: 0.7180 - val_loss: 16.8772 - val_accuracy: 0.5126\n",
      "Epoch 46/50\n",
      "119997/119997 [==============================] - 5s 40us/sample - loss: 0.5161 - accuracy: 0.7120 - val_loss: 13.7438 - val_accuracy: 0.5380\n",
      "Epoch 47/50\n",
      "119997/119997 [==============================] - 5s 44us/sample - loss: 0.5018 - accuracy: 0.7173 - val_loss: 32.6855 - val_accuracy: 0.5211\n",
      "Epoch 48/50\n",
      "119997/119997 [==============================] - 5s 43us/sample - loss: 0.4755 - accuracy: 0.7498 - val_loss: 32.8847 - val_accuracy: 0.5192\n",
      "Epoch 49/50\n",
      "119997/119997 [==============================] - 5s 39us/sample - loss: 0.4641 - accuracy: 0.7609 - val_loss: 17.3161 - val_accuracy: 0.5238\n",
      "Epoch 50/50\n",
      "119997/119997 [==============================] - 5s 39us/sample - loss: 0.4575 - accuracy: 0.7666 - val_loss: 17.9562 - val_accuracy: 0.5289\n",
      "3999/3999 [==============================] - 0s 17us/sample - loss: 6.6872 - accuracy: 0.5039\n",
      "3999/3999 [==============================] - 0s 17us/sample - loss: 17.8778 - accuracy: 0.5626\n",
      "3999/3999 [==============================] - 0s 16us/sample - loss: 12.6808 - accuracy: 0.6547\n",
      "6-layers-64-nodes\n",
      "Train on 119997 samples, validate on 29997 samples\n",
      "Epoch 1/50\n",
      "119997/119997 [==============================] - 6s 50us/sample - loss: 0.6910 - accuracy: 0.5016 - val_loss: 0.6799 - val_accuracy: 0.4999\n",
      "Epoch 2/50\n",
      "119997/119997 [==============================] - 6s 47us/sample - loss: 0.6898 - accuracy: 0.5054 - val_loss: 0.6764 - val_accuracy: 0.4999\n",
      "Epoch 3/50\n",
      "119997/119997 [==============================] - 6s 47us/sample - loss: 0.6896 - accuracy: 0.5049 - val_loss: 0.6906 - val_accuracy: 0.4778\n",
      "Epoch 4/50\n",
      "119997/119997 [==============================] - 5s 42us/sample - loss: 0.6896 - accuracy: 0.5040 - val_loss: 0.6755 - val_accuracy: 0.5267\n",
      "Epoch 5/50\n",
      "119997/119997 [==============================] - 5s 45us/sample - loss: 0.6897 - accuracy: 0.5048 - val_loss: 0.6736 - val_accuracy: 0.5291\n",
      "Epoch 6/50\n",
      "119997/119997 [==============================] - 5s 42us/sample - loss: 0.6891 - accuracy: 0.5050 - val_loss: 0.6892 - val_accuracy: 0.5419\n",
      "Epoch 7/50\n",
      "119997/119997 [==============================] - 5s 41us/sample - loss: 0.6889 - accuracy: 0.5066 - val_loss: 0.6788 - val_accuracy: 0.5282\n",
      "Epoch 8/50\n",
      "119997/119997 [==============================] - 6s 47us/sample - loss: 0.6887 - accuracy: 0.5058 - val_loss: 0.6737 - val_accuracy: 0.5351\n",
      "Epoch 9/50\n",
      "119997/119997 [==============================] - 5s 44us/sample - loss: 0.6887 - accuracy: 0.5067 - val_loss: 0.6829 - val_accuracy: 0.5293\n",
      "Epoch 10/50\n",
      "119997/119997 [==============================] - 6s 51us/sample - loss: 0.6884 - accuracy: 0.5080 - val_loss: 0.6940 - val_accuracy: 0.4982\n",
      "Epoch 11/50\n",
      "119997/119997 [==============================] - 6s 50us/sample - loss: 0.6884 - accuracy: 0.5049 - val_loss: 0.6754 - val_accuracy: 0.5380\n",
      "Epoch 12/50\n",
      "119997/119997 [==============================] - 5s 45us/sample - loss: 0.6883 - accuracy: 0.5072 - val_loss: 0.6986 - val_accuracy: 0.4926\n",
      "Epoch 13/50\n",
      "119997/119997 [==============================] - 6s 48us/sample - loss: 0.6883 - accuracy: 0.5079 - val_loss: 0.6861 - val_accuracy: 0.5027\n",
      "Epoch 14/50\n",
      "119997/119997 [==============================] - 6s 48us/sample - loss: 0.6884 - accuracy: 0.5070 - val_loss: 0.6847 - val_accuracy: 0.5353\n",
      "Epoch 15/50\n",
      "119997/119997 [==============================] - 5s 45us/sample - loss: 0.6884 - accuracy: 0.5070 - val_loss: 0.7214 - val_accuracy: 0.5127\n",
      "Epoch 16/50\n",
      "119997/119997 [==============================] - 6s 47us/sample - loss: 0.6881 - accuracy: 0.5089 - val_loss: 0.6787 - val_accuracy: 0.5164\n",
      "Epoch 17/50\n",
      "119997/119997 [==============================] - 6s 46us/sample - loss: 0.6883 - accuracy: 0.5120 - val_loss: 0.6950 - val_accuracy: 0.5066\n",
      "Epoch 18/50\n",
      "119997/119997 [==============================] - 5s 42us/sample - loss: 0.6881 - accuracy: 0.5134 - val_loss: 0.7012 - val_accuracy: 0.5334\n",
      "Epoch 19/50\n",
      "119997/119997 [==============================] - 6s 46us/sample - loss: 0.6850 - accuracy: 0.5207 - val_loss: 0.6718 - val_accuracy: 0.5380\n",
      "Epoch 20/50\n",
      "119997/119997 [==============================] - 5s 44us/sample - loss: 0.6801 - accuracy: 0.5261 - val_loss: 0.6812 - val_accuracy: 0.5434\n",
      "Epoch 21/50\n",
      "119997/119997 [==============================] - 5s 45us/sample - loss: 0.6731 - accuracy: 0.5315 - val_loss: 0.8563 - val_accuracy: 0.5535\n",
      "Epoch 22/50\n",
      "119997/119997 [==============================] - 6s 47us/sample - loss: 0.6660 - accuracy: 0.5388 - val_loss: 0.8285 - val_accuracy: 0.5385\n",
      "Epoch 23/50\n",
      "119997/119997 [==============================] - 5s 42us/sample - loss: 0.6606 - accuracy: 0.5406 - val_loss: 0.8386 - val_accuracy: 0.5238\n",
      "Epoch 24/50\n",
      "119997/119997 [==============================] - 5s 43us/sample - loss: 0.6604 - accuracy: 0.5420 - val_loss: 1.0016 - val_accuracy: 0.5420\n",
      "Epoch 25/50\n",
      "119997/119997 [==============================] - 6s 49us/sample - loss: 0.6557 - accuracy: 0.5457 - val_loss: 0.8839 - val_accuracy: 0.5548\n",
      "Epoch 26/50\n",
      "119997/119997 [==============================] - 5s 45us/sample - loss: 0.6555 - accuracy: 0.5467 - val_loss: 1.0033 - val_accuracy: 0.4915\n",
      "Epoch 27/50\n",
      "119997/119997 [==============================] - 6s 51us/sample - loss: 0.6570 - accuracy: 0.5457 - val_loss: 0.8034 - val_accuracy: 0.5266\n",
      "Epoch 28/50\n",
      "119997/119997 [==============================] - 5s 42us/sample - loss: 0.6582 - accuracy: 0.5435 - val_loss: 0.8489 - val_accuracy: 0.5140\n",
      "Epoch 29/50\n",
      "119997/119997 [==============================] - 5s 42us/sample - loss: 0.6586 - accuracy: 0.5440 - val_loss: 0.9159 - val_accuracy: 0.5435\n",
      "Epoch 30/50\n",
      "119997/119997 [==============================] - 6s 47us/sample - loss: 0.6562 - accuracy: 0.5453 - val_loss: 0.9272 - val_accuracy: 0.4778\n",
      "Epoch 31/50\n",
      "119997/119997 [==============================] - 5s 42us/sample - loss: 0.6514 - accuracy: 0.5518 - val_loss: 2.2155 - val_accuracy: 0.4871\n",
      "Epoch 32/50\n",
      "119997/119997 [==============================] - 5s 43us/sample - loss: 0.6492 - accuracy: 0.5578 - val_loss: 1.5268 - val_accuracy: 0.4865\n",
      "Epoch 33/50\n",
      "119997/119997 [==============================] - 6s 47us/sample - loss: 0.6501 - accuracy: 0.5545 - val_loss: 0.9961 - val_accuracy: 0.4872\n",
      "Epoch 34/50\n",
      "119997/119997 [==============================] - 5s 43us/sample - loss: 0.6517 - accuracy: 0.5539 - val_loss: 1.5780 - val_accuracy: 0.4883\n",
      "Epoch 35/50\n",
      "119997/119997 [==============================] - 5s 45us/sample - loss: 0.6450 - accuracy: 0.5604 - val_loss: 1.0312 - val_accuracy: 0.4885\n",
      "Epoch 36/50\n",
      "119997/119997 [==============================] - 6s 47us/sample - loss: 0.6518 - accuracy: 0.5555 - val_loss: 2.2372 - val_accuracy: 0.4882\n",
      "Epoch 37/50\n",
      "119997/119997 [==============================] - 6s 46us/sample - loss: 0.6437 - accuracy: 0.5688 - val_loss: 1.6323 - val_accuracy: 0.4883\n",
      "Epoch 38/50\n",
      "119997/119997 [==============================] - 5s 42us/sample - loss: 0.6244 - accuracy: 0.5907 - val_loss: 2.2186 - val_accuracy: 0.4965\n",
      "Epoch 39/50\n",
      "119997/119997 [==============================] - 5s 42us/sample - loss: 0.6114 - accuracy: 0.6051 - val_loss: 5.4765 - val_accuracy: 0.5117\n",
      "Epoch 40/50\n",
      "119997/119997 [==============================] - 5s 42us/sample - loss: 0.5561 - accuracy: 0.6657 - val_loss: 10.3987 - val_accuracy: 0.5471\n",
      "Epoch 41/50\n",
      "119997/119997 [==============================] - 5s 45us/sample - loss: 0.5346 - accuracy: 0.6903 - val_loss: 16.3612 - val_accuracy: 0.5385\n",
      "Epoch 42/50\n",
      "119997/119997 [==============================] - 5s 43us/sample - loss: 0.4899 - accuracy: 0.7405 - val_loss: 15.3144 - val_accuracy: 0.5590\n",
      "Epoch 43/50\n",
      "119997/119997 [==============================] - 5s 41us/sample - loss: 0.4802 - accuracy: 0.7474 - val_loss: 19.9212 - val_accuracy: 0.5576\n",
      "Epoch 44/50\n",
      "119997/119997 [==============================] - 5s 42us/sample - loss: 0.5183 - accuracy: 0.7194 - val_loss: 17.3009 - val_accuracy: 0.5542\n",
      "Epoch 45/50\n",
      "119997/119997 [==============================] - 5s 41us/sample - loss: 0.4854 - accuracy: 0.7410 - val_loss: 20.5155 - val_accuracy: 0.5280\n",
      "Epoch 46/50\n",
      "119997/119997 [==============================] - 5s 41us/sample - loss: 0.4633 - accuracy: 0.7645 - val_loss: 26.9363 - val_accuracy: 0.5557\n",
      "Epoch 47/50\n",
      "119997/119997 [==============================] - 5s 45us/sample - loss: 0.4688 - accuracy: 0.7575 - val_loss: 35.7169 - val_accuracy: 0.5601\n",
      "Epoch 48/50\n",
      "119997/119997 [==============================] - 5s 42us/sample - loss: 0.4651 - accuracy: 0.7636 - val_loss: 36.8370 - val_accuracy: 0.5396\n",
      "Epoch 49/50\n",
      "119997/119997 [==============================] - 5s 41us/sample - loss: 0.4532 - accuracy: 0.7695 - val_loss: 29.1454 - val_accuracy: 0.5572\n",
      "Epoch 50/50\n",
      "119997/119997 [==============================] - 5s 45us/sample - loss: 0.4605 - accuracy: 0.7630 - val_loss: 34.4289 - val_accuracy: 0.5612\n",
      "3999/3999 [==============================] - 0s 21us/sample - loss: 15.4694 - accuracy: 0.4241\n",
      "3999/3999 [==============================] - 0s 21us/sample - loss: 1.5662 - accuracy: 0.6044\n",
      "3999/3999 [==============================] - 0s 20us/sample - loss: 0.6568 - accuracy: 0.7697\n",
      "6-layers-128-nodes\n",
      "Train on 119997 samples, validate on 29997 samples\n",
      "Epoch 1/50\n",
      "119997/119997 [==============================] - 7s 57us/sample - loss: 0.6912 - accuracy: 0.5023 - val_loss: 0.6742 - val_accuracy: 0.5289\n",
      "Epoch 2/50\n",
      "119997/119997 [==============================] - 7s 56us/sample - loss: 0.6899 - accuracy: 0.5041 - val_loss: 0.6708 - val_accuracy: 0.4999\n",
      "Epoch 3/50\n",
      "119997/119997 [==============================] - 6s 53us/sample - loss: 0.6900 - accuracy: 0.5044 - val_loss: 0.6763 - val_accuracy: 0.5247\n",
      "Epoch 4/50\n",
      "119997/119997 [==============================] - 6s 52us/sample - loss: 0.6898 - accuracy: 0.5029 - val_loss: 0.6753 - val_accuracy: 0.5265\n",
      "Epoch 5/50\n",
      "119997/119997 [==============================] - 6s 52us/sample - loss: 0.6899 - accuracy: 0.5042 - val_loss: 0.7956 - val_accuracy: 0.5826\n",
      "Epoch 6/50\n",
      "119997/119997 [==============================] - 6s 51us/sample - loss: 0.6900 - accuracy: 0.5031 - val_loss: 0.6702 - val_accuracy: 0.5609\n",
      "Epoch 7/50\n",
      "119997/119997 [==============================] - 7s 56us/sample - loss: 0.6896 - accuracy: 0.5037 - val_loss: 0.6756 - val_accuracy: 0.5261\n",
      "Epoch 8/50\n",
      "119997/119997 [==============================] - 6s 52us/sample - loss: 0.6895 - accuracy: 0.5057 - val_loss: 0.6787 - val_accuracy: 0.5438\n",
      "Epoch 9/50\n",
      "119997/119997 [==============================] - 6s 52us/sample - loss: 0.6896 - accuracy: 0.5043 - val_loss: 0.6698 - val_accuracy: 0.5360\n",
      "Epoch 10/50\n",
      "119997/119997 [==============================] - 6s 52us/sample - loss: 0.6896 - accuracy: 0.5039 - val_loss: 0.6706 - val_accuracy: 0.5335\n",
      "Epoch 11/50\n",
      "119997/119997 [==============================] - 6s 52us/sample - loss: 0.6895 - accuracy: 0.5031 - val_loss: 0.6791 - val_accuracy: 0.5203\n",
      "Epoch 12/50\n",
      "119997/119997 [==============================] - 7s 56us/sample - loss: 0.6896 - accuracy: 0.5036 - val_loss: 0.6727 - val_accuracy: 0.5323\n",
      "Epoch 13/50\n",
      "119997/119997 [==============================] - 7s 55us/sample - loss: 0.6890 - accuracy: 0.5063 - val_loss: 0.6704 - val_accuracy: 0.5355\n",
      "Epoch 14/50\n",
      "119997/119997 [==============================] - 6s 52us/sample - loss: 0.6895 - accuracy: 0.5062 - val_loss: 0.6713 - val_accuracy: 0.4999\n",
      "Epoch 15/50\n",
      "119997/119997 [==============================] - 6s 52us/sample - loss: 0.6893 - accuracy: 0.5035 - val_loss: 0.6937 - val_accuracy: 0.5451\n",
      "Epoch 16/50\n",
      "119997/119997 [==============================] - 7s 54us/sample - loss: 0.6899 - accuracy: 0.5050 - val_loss: 0.6735 - val_accuracy: 0.4999\n",
      "Epoch 17/50\n",
      "119997/119997 [==============================] - 6s 53us/sample - loss: 0.6893 - accuracy: 0.5028 - val_loss: 0.6709 - val_accuracy: 0.5334\n",
      "Epoch 18/50\n",
      "119997/119997 [==============================] - 6s 51us/sample - loss: 0.6891 - accuracy: 0.5039 - val_loss: 0.6712 - val_accuracy: 0.5320\n",
      "Epoch 19/50\n",
      "119997/119997 [==============================] - 7s 55us/sample - loss: 0.6892 - accuracy: 0.5050 - val_loss: 0.6745 - val_accuracy: 0.4999\n",
      "Epoch 20/50\n",
      "119997/119997 [==============================] - 6s 53us/sample - loss: 0.6892 - accuracy: 0.5037 - val_loss: 0.6724 - val_accuracy: 0.5305\n",
      "Epoch 21/50\n",
      "119997/119997 [==============================] - 7s 56us/sample - loss: 0.6891 - accuracy: 0.5055 - val_loss: 0.6715 - val_accuracy: 0.5321\n",
      "Epoch 22/50\n",
      "119997/119997 [==============================] - 6s 52us/sample - loss: 0.6892 - accuracy: 0.5045 - val_loss: 0.6682 - val_accuracy: 0.5392\n",
      "Epoch 23/50\n",
      "119997/119997 [==============================] - 6s 52us/sample - loss: 0.6891 - accuracy: 0.5043 - val_loss: 0.6717 - val_accuracy: 0.5365\n",
      "Epoch 24/50\n",
      "119997/119997 [==============================] - 6s 52us/sample - loss: 0.6896 - accuracy: 0.5044 - val_loss: 0.6725 - val_accuracy: 0.5329\n",
      "Epoch 25/50\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "119997/119997 [==============================] - 6s 52us/sample - loss: 0.6893 - accuracy: 0.5047 - val_loss: 0.6658 - val_accuracy: 0.5459\n",
      "Epoch 26/50\n",
      "119997/119997 [==============================] - 7s 55us/sample - loss: 0.6892 - accuracy: 0.5045 - val_loss: 0.6721 - val_accuracy: 0.5315\n",
      "Epoch 27/50\n",
      "119997/119997 [==============================] - 6s 52us/sample - loss: 0.6892 - accuracy: 0.5051 - val_loss: 0.6792 - val_accuracy: 0.5387\n",
      "Epoch 28/50\n",
      "119997/119997 [==============================] - 6s 53us/sample - loss: 0.6894 - accuracy: 0.5057 - val_loss: 0.6748 - val_accuracy: 0.5306\n",
      "Epoch 29/50\n",
      "119997/119997 [==============================] - 6s 52us/sample - loss: 0.6892 - accuracy: 0.5054 - val_loss: 0.6787 - val_accuracy: 0.4987\n",
      "Epoch 30/50\n",
      "119997/119997 [==============================] - 6s 51us/sample - loss: 0.6893 - accuracy: 0.5054 - val_loss: 0.6739 - val_accuracy: 0.5295\n",
      "Epoch 31/50\n",
      "119997/119997 [==============================] - 7s 56us/sample - loss: 0.6890 - accuracy: 0.5052 - val_loss: 0.6753 - val_accuracy: 0.5358\n",
      "Epoch 32/50\n",
      "119997/119997 [==============================] - 7s 60us/sample - loss: 0.6893 - accuracy: 0.5036 - val_loss: 0.6769 - val_accuracy: 0.5427\n",
      "Epoch 33/50\n",
      "119997/119997 [==============================] - 8s 67us/sample - loss: 0.6890 - accuracy: 0.5045 - val_loss: 0.6765 - val_accuracy: 0.5349\n",
      "Epoch 34/50\n",
      "119997/119997 [==============================] - 7s 54us/sample - loss: 0.6893 - accuracy: 0.5060 - val_loss: 0.6732 - val_accuracy: 0.5309\n",
      "Epoch 35/50\n",
      "119997/119997 [==============================] - 7s 62us/sample - loss: 0.6891 - accuracy: 0.5061 - val_loss: 0.6723 - val_accuracy: 0.5348\n",
      "Epoch 36/50\n",
      "119997/119997 [==============================] - 6s 53us/sample - loss: 0.6891 - accuracy: 0.5047 - val_loss: 0.6714 - val_accuracy: 0.5334\n",
      "Epoch 37/50\n",
      "119997/119997 [==============================] - 7s 56us/sample - loss: 0.6889 - accuracy: 0.5047 - val_loss: 0.6741 - val_accuracy: 0.5294\n",
      "Epoch 38/50\n",
      "119997/119997 [==============================] - 6s 53us/sample - loss: 0.6897 - accuracy: 0.5049 - val_loss: 0.6741 - val_accuracy: 0.5305\n",
      "Epoch 39/50\n",
      "119997/119997 [==============================] - 7s 62us/sample - loss: 0.6894 - accuracy: 0.5058 - val_loss: 0.6743 - val_accuracy: 0.5282\n",
      "Epoch 40/50\n",
      "119997/119997 [==============================] - 6s 53us/sample - loss: 0.6892 - accuracy: 0.5049 - val_loss: 0.6752 - val_accuracy: 0.5291\n",
      "Epoch 41/50\n",
      "119997/119997 [==============================] - 6s 53us/sample - loss: 0.6892 - accuracy: 0.5055 - val_loss: 0.6773 - val_accuracy: 0.5333\n",
      "Epoch 42/50\n",
      "119997/119997 [==============================] - 6s 54us/sample - loss: 0.6890 - accuracy: 0.5050 - val_loss: 0.6741 - val_accuracy: 0.5316\n",
      "Epoch 43/50\n",
      "119997/119997 [==============================] - 6s 53us/sample - loss: 0.6890 - accuracy: 0.5061 - val_loss: 0.6852 - val_accuracy: 0.5363\n",
      "Epoch 44/50\n",
      "119997/119997 [==============================] - 7s 62us/sample - loss: 0.6893 - accuracy: 0.5057 - val_loss: 0.6745 - val_accuracy: 0.4999\n",
      "Epoch 45/50\n",
      "119997/119997 [==============================] - 6s 52us/sample - loss: 0.6895 - accuracy: 0.5041 - val_loss: 0.6743 - val_accuracy: 0.5290\n",
      "Epoch 46/50\n",
      "119997/119997 [==============================] - 6s 53us/sample - loss: 0.6891 - accuracy: 0.5047 - val_loss: 0.6797 - val_accuracy: 0.5341\n",
      "Epoch 47/50\n",
      "119997/119997 [==============================] - 6s 53us/sample - loss: 0.6890 - accuracy: 0.5067 - val_loss: 0.6705 - val_accuracy: 0.5369\n",
      "Epoch 48/50\n",
      "119997/119997 [==============================] - 7s 57us/sample - loss: 0.6887 - accuracy: 0.5049 - val_loss: 0.6718 - val_accuracy: 0.5345\n",
      "Epoch 49/50\n",
      "119997/119997 [==============================] - 7s 58us/sample - loss: 0.6887 - accuracy: 0.5050 - val_loss: 0.6723 - val_accuracy: 0.5337\n",
      "Epoch 50/50\n",
      "119997/119997 [==============================] - 6s 52us/sample - loss: 0.6888 - accuracy: 0.5050 - val_loss: 0.6732 - val_accuracy: 0.5361\n",
      "3999/3999 [==============================] - 0s 25us/sample - loss: 0.7933 - accuracy: 0.4846\n",
      "3999/3999 [==============================] - 0s 21us/sample - loss: 0.6454 - accuracy: 0.5724\n",
      "3999/3999 [==============================] - 0s 25us/sample - loss: 0.5967 - accuracy: 0.6412\n"
     ]
    }
   ],
   "source": [
    "num_layers = [2,3,4,5,6]\n",
    "layer_sizes = [16,32,64,128]\n",
    "\n",
    "for layer in num_layers:\n",
    "    for layer_size in layer_sizes:\n",
    "        NAME = \"{}-layers-{}-nodes\".format(layer, layer_size)\n",
    "        LOGNAME = \"{}-layers-{}-nodes-{}\".format(layer, layer_size, int(time.time()))\n",
    "        tensorboard = TensorBoard(log_dir='logs\\{}'.format(LOGNAME))\n",
    "        print(NAME)\n",
    "        \n",
    "        model = keras.Sequential()\n",
    "        model.add(tf.keras.Input(shape=(numFeatures,)))\n",
    "        \n",
    "        ##Adding Layers\n",
    "        for i in range(layer):\n",
    "            model.add(  tf.keras.layers.Dense(layer_size,activation='relu')  )\n",
    "        \n",
    "        ##Output Layer\n",
    "        model.add( tf.keras.layers.Dense(1,activation='sigmoid') )\n",
    "        \n",
    "        ##Compiling model\n",
    "        model.compile(loss='binary_crossentropy',\n",
    "                     optimizer='adam',\n",
    "                     metrics = ['accuracy'])\n",
    "        \n",
    "        ##Training Model\n",
    "        model.fit(X,y, batch_size = 32,\n",
    "                  epochs = NUM_EPOCHS,\n",
    "                  validation_data=(Xv,yv),\n",
    "                  callbacks = [tensorboard] )\n",
    "        \n",
    "        ##Saving Model\n",
    "        model.save('SavedModels\\{}.h5'.format(NAME))\n",
    "        \n",
    "        ##Testing Model\n",
    "        results1 = model.evaluate(Xt1, yt1)\n",
    "        results2 = model.evaluate(Xt2, yt2)\n",
    "        results3 = model.evaluate(Xt3, yt3)\n",
    "        \n",
    "        resultsLow.append(results1)\n",
    "        resultsMid.append(results2)\n",
    "        resultsHigh.append(results3)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0\n",
      "1\n"
     ]
    }
   ],
   "source": [
    "for i in range(2):\n",
    "    print(i)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0\n"
     ]
    }
   ],
   "source": [
    "for i in range(1):print(i)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "for i"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.6.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}