Run RNN tests

RNN_1.2A.py

@@ -0,0 +1,26 @@
+
+from keras.models import Sequential
+from keras.layers import Dense, Activation, SimpleRNN
+from keras.utils.vis_utils import plot_model
+import os
+import numpy as np
+from reader import Reader
+
+
+length = Reader.getInputShape()
+
+model = Sequential()
+
+#EXPECTS INPUT AS (nb_sample, timesteps, nb_features), where nb_sample=1 (batch_size = 1), timesteps = 1 and nb_features = length
+
+#model.add(Dense(40, input_dim = 12, init='uniform', activation='relu'))
+model.add(SimpleRNN(units=50, input_shape=(1,length), batch_input_shape=(1,1,length), recurrent_initializer='random_uniform', kernel_initializer='random_uniform', activation='sigmoid', stateful=True))
+model.add(Dense(1, kernel_initializer='random_uniform', activation = 'sigmoid'))
+
+
+model.summary()
+plot_model(model, to_file='./RNNmodels/RNN_1.2A.png', show_shapes=True)
+fp = open('./RNNmodels/RNN_1.2A.json', 'w')
+fp.write(model.to_json())
+fp.close()
+

RNN_2.2A.py

@@ -0,0 +1,27 @@
+
+from keras.models import Sequential
+from keras.layers import Dense, Activation, SimpleRNN
+from keras.utils.vis_utils import plot_model
+import os
+import numpy as np
+from reader import Reader
+
+
+length = Reader.getInputShape()
+
+model = Sequential()
+
+#EXPECTS INPUT AS (nb_sample, timesteps, nb_features), where nb_sample=1 (batch_size = 1), timesteps = 1 and nb_features = length
+
+#model.add(Dense(40, input_dim = 12, init='uniform', activation='relu'))
+model.add(SimpleRNN(units=50, input_shape=(1,length), batch_input_shape=(1,1,length), recurrent_initializer='random_uniform', kernel_initializer='random_uniform', activation='sigmoid', return_sequences=True, stateful=True))
+model.add(SimpleRNN(units=50, recurrent_initializer='random_uniform', kernel_initializer='random_uniform', activation='sigmoid', stateful=True))
+model.add(Dense(1, kernel_initializer='random_uniform', activation = 'sigmoid'))
+
+
+model.summary()
+plot_model(model, to_file='./RNNmodels/RNN_2.2A.png', show_shapes=True)
+fp = open('./RNNmodels/RNN_2.2A.json', 'w')
+fp.write(model.to_json())
+fp.close()
+

RNN_3.2A.py

@@ -0,0 +1,28 @@
+
+from keras.models import Sequential
+from keras.layers import Dense, Activation, SimpleRNN
+from keras.utils.vis_utils import plot_model
+import os
+import numpy as np
+from reader import Reader
+
+
+length = Reader.getInputShape()
+
+model = Sequential()
+
+#EXPECTS INPUT AS (nb_sample, timesteps, nb_features), where nb_sample=1 (batch_size = 1), timesteps = 1 and nb_features = length
+
+#model.add(Dense(40, input_dim = 12, init='uniform', activation='relu'))
+model.add(SimpleRNN(units=50, input_shape=(1,length), batch_input_shape=(1,1,length), recurrent_initializer='random_uniform', kernel_initializer='random_uniform', activation='sigmoid', return_sequences=True, stateful=True))
+model.add(SimpleRNN(units=40, recurrent_initializer='random_uniform', kernel_initializer='random_uniform', activation='sigmoid', stateful=True, return_sequences=True))
+model.add(SimpleRNN(units=30, recurrent_initializer='random_uniform', kernel_initializer='random_uniform', activation='sigmoid', stateful=True))
+model.add(Dense(1, kernel_initializer='random_uniform', activation = 'sigmoid'))
+
+
+model.summary()
+plot_model(model, to_file='./RNNmodels/RNN_3.2A.png', show_shapes=True)
+fp = open('./RNNmodels/RNN_3.2A.json', 'w')
+fp.write(model.to_json())
+fp.close()
+

RNN_plot_all.py

@@ -22,8 +22,8 @@
 #(4, 'Dataset4', 2, 'RNN_1A') !!! mixed type binary and continous
 #(5, 'Dataset5', 3, 'RNN_1B') +categorical
 #(6, 'Dataset0', 5, 'RNN_1C') +categorical
-dataset_id = 4
-dataset_name = 'Dataset4'
+dataset_id = 3
+dataset_name = 'Dataset3'
 num_classes = 2
 #RNN_name = 'RNN_1A'
 
@@ -35,9 +35,9 @@
 #x_test = x_test[0:1000,:]
 #y_test = y_test[0:1000]
 
-for RNN_name in ['RNN_3A']:
+for RNN_name in ['RNN_1.2A', 'RNN_2.2A', 'RNN_3.2A']:
 	results = {}
-	for loss,optimizer in [('mse','sgd')]:#, ('binary_crossentropy','rmsprop')]:	#categorical_crossentropy
+	for loss,optimizer in [('binary_crossentropy','rmsprop')]:#, ('binary_crossentropy','rmsprop')]:	#categorical_crossentropy
 	#for optimizer in ['sgd', 'rmsprop']:
 		RNNmodel = RNNHandler.RNNHandler(RNN_name, num_classes, loss, optimizer)
 		(res_loss, res_accuracy, res_precision, res_recall, res_fscore) = RNNmodel.fit_and_eval(x_train, y_train, x_test, y_test, num_epochs, dataset_name)

RNNmodels/RNN_1.2A.json

@@ -0,0 +1 @@
+{"backend": "theano", "config": [{"class_name": "SimpleRNN", "config": {"dropout": 0.0, "use_bias": true, "activation": "sigmoid", "recurrent_regularizer": null, "dtype": "float32", "batch_input_shape": [1, 1, 79], "stateful": true, "recurrent_initializer": {"class_name": "RandomUniform", "config": {"maxval": 0.05, "minval": -0.05, "seed": null}}, "bias_regularizer": null, "units": 50, "activity_regularizer": null, "recurrent_constraint": null, "kernel_initializer": {"class_name": "RandomUniform", "config": {"maxval": 0.05, "minval": -0.05, "seed": null}}, "bias_initializer": {"class_name": "Zeros", "config": {}}, "recurrent_dropout": 0.0, "unroll": false, "kernel_regularizer": null, "kernel_constraint": null, "return_sequences": false, "implementation": 0, "trainable": true, "name": "simple_rnn_1", "go_backwards": false, "bias_constraint": null}}, {"class_name": "Dense", "config": {"use_bias": true, "bias_initializer": {"class_name": "Zeros", "config": {}}, "kernel_regularizer": null, "activation": "sigmoid", "kernel_constraint": null, "kernel_initializer": {"class_name": "RandomUniform", "config": {"maxval": 0.05, "minval": -0.05, "seed": null}}, "units": 1, "activity_regularizer": null, "trainable": true, "name": "dense_1", "bias_regularizer": null, "bias_constraint": null}}], "keras_version": "2.0.3", "class_name": "Sequential"}

RNNmodels/RNN_2.2A.json

@@ -0,0 +1 @@
+{"class_name": "Sequential", "config": [{"class_name": "SimpleRNN", "config": {"bias_initializer": {"class_name": "Zeros", "config": {}}, "dtype": "float32", "dropout": 0.0, "implementation": 0, "batch_input_shape": [1, 1, 79], "kernel_initializer": {"class_name": "RandomUniform", "config": {"seed": null, "minval": -0.05, "maxval": 0.05}}, "trainable": true, "activation": "sigmoid", "bias_regularizer": null, "units": 50, "unroll": false, "stateful": true, "kernel_regularizer": null, "name": "simple_rnn_1", "bias_constraint": null, "kernel_constraint": null, "recurrent_dropout": 0.0, "recurrent_initializer": {"class_name": "RandomUniform", "config": {"seed": null, "minval": -0.05, "maxval": 0.05}}, "go_backwards": false, "return_sequences": true, "recurrent_constraint": null, "use_bias": true, "activity_regularizer": null, "recurrent_regularizer": null}}, {"class_name": "SimpleRNN", "config": {"bias_initializer": {"class_name": "Zeros", "config": {}}, "dropout": 0.0, "implementation": 0, "kernel_initializer": {"class_name": "RandomUniform", "config": {"seed": null, "minval": -0.05, "maxval": 0.05}}, "use_bias": true, "trainable": true, "activation": "sigmoid", "bias_regularizer": null, "units": 50, "unroll": false, "stateful": true, "kernel_regularizer": null, "name": "simple_rnn_2", "bias_constraint": null, "kernel_constraint": null, "recurrent_initializer": {"class_name": "RandomUniform", "config": {"seed": null, "minval": -0.05, "maxval": 0.05}}, "go_backwards": false, "return_sequences": false, "recurrent_constraint": null, "recurrent_regularizer": null, "activity_regularizer": null, "recurrent_dropout": 0.0}}, {"class_name": "Dense", "config": {"kernel_regularizer": null, "name": "dense_1", "bias_constraint": null, "kernel_initializer": {"class_name": "RandomUniform", "config": {"seed": null, "minval": -0.05, "maxval": 0.05}}, "kernel_constraint": null, "bias_initializer": {"class_name": "Zeros", "config": {}}, "trainable": true, "activation": "sigmoid", "bias_regularizer": null, "use_bias": true, "units": 1, "activity_regularizer": null}}], "backend": "theano", "keras_version": "2.0.3"}

RNNmodels/RNN_3.2A.json

@@ -0,0 +1 @@
+{"config": [{"config": {"stateful": true, "units": 50, "trainable": true, "bias_regularizer": null, "implementation": 0, "kernel_regularizer": null, "kernel_constraint": null, "go_backwards": false, "batch_input_shape": [1, 1, 79], "dropout": 0.0, "recurrent_constraint": null, "activity_regularizer": null, "use_bias": true, "unroll": false, "bias_constraint": null, "kernel_initializer": {"config": {"minval": -0.05, "seed": null, "maxval": 0.05}, "class_name": "RandomUniform"}, "activation": "sigmoid", "bias_initializer": {"config": {}, "class_name": "Zeros"}, "recurrent_regularizer": null, "return_sequences": true, "recurrent_dropout": 0.0, "name": "simple_rnn_1", "recurrent_initializer": {"config": {"minval": -0.05, "seed": null, "maxval": 0.05}, "class_name": "RandomUniform"}, "dtype": "float32"}, "class_name": "SimpleRNN"}, {"config": {"stateful": true, "units": 40, "trainable": true, "bias_regularizer": null, "implementation": 0, "go_backwards": false, "kernel_constraint": null, "kernel_regularizer": null, "name": "simple_rnn_2", "recurrent_constraint": null, "activity_regularizer": null, "use_bias": true, "unroll": false, "bias_constraint": null, "kernel_initializer": {"config": {"minval": -0.05, "seed": null, "maxval": 0.05}, "class_name": "RandomUniform"}, "activation": "sigmoid", "recurrent_dropout": 0.0, "recurrent_regularizer": null, "return_sequences": true, "bias_initializer": {"config": {}, "class_name": "Zeros"}, "dropout": 0.0, "recurrent_initializer": {"config": {"minval": -0.05, "seed": null, "maxval": 0.05}, "class_name": "RandomUniform"}}, "class_name": "SimpleRNN"}, {"config": {"stateful": true, "units": 30, "trainable": true, "bias_regularizer": null, "implementation": 0, "go_backwards": false, "kernel_constraint": null, "kernel_regularizer": null, "name": "simple_rnn_3", "recurrent_constraint": null, "activity_regularizer": null, "use_bias": true, "unroll": false, "bias_constraint": null, "kernel_initializer": {"config": {"minval": -0.05, "seed": null, "maxval": 0.05}, "class_name": "RandomUniform"}, "activation": "sigmoid", "recurrent_dropout": 0.0, "recurrent_regularizer": null, "return_sequences": false, "bias_initializer": {"config": {}, "class_name": "Zeros"}, "dropout": 0.0, "recurrent_initializer": {"config": {"minval": -0.05, "seed": null, "maxval": 0.05}, "class_name": "RandomUniform"}}, "class_name": "SimpleRNN"}, {"config": {"units": 1, "activity_regularizer": null, "trainable": true, "use_bias": true, "bias_regularizer": null, "kernel_initializer": {"config": {"minval": -0.05, "seed": null, "maxval": 0.05}, "class_name": "RandomUniform"}, "activation": "sigmoid", "kernel_regularizer": null, "kernel_constraint": null, "bias_initializer": {"config": {}, "class_name": "Zeros"}, "name": "dense_1", "bias_constraint": null}, "class_name": "Dense"}], "keras_version": "2.0.3", "backend": "theano", "class_name": "Sequential"}

RNNresults/X.Dataset3.RNN_2A_rmsprop

@@ -0,0 +1,12 @@
+RNN_2.2A Dataset3 Loss:binary_crossentropy Optimizer:rmsprop Dropout:No
+Epoch|Loss|Accuracy|Precision|Recall|Fscore
+1|4.11098907206|0.744833333333|0.986037586024|0.4967|0.660622450789
+2|4.11092153752|0.744905555556|0.986148790225|0.496788888889|0.660726028713
+3|4.11085149486|0.744916666667|0.98619229327|0.496788888889|0.660735792873
+4|4.11084888653|0.744916666667|0.98619229327|0.496788888889|0.660735792873
+5|4.11084704866|0.744916666667|0.98619229327|0.496788888889|0.660735792873
+6|4.11084670704|0.744916666667|0.98619229327|0.496788888889|0.660735792873
+7|4.11084650289|0.744916666667|0.98619229327|0.496788888889|0.660735792873
+8|4.11084656372|0.744916666667|0.98619229327|0.496788888889|0.660735792873
+9|4.11084650949|0.744916666667|0.98619229327|0.496788888889|0.660735792873
+10|4.11084649933|0.744916666667|0.98619229327|0.496788888889|0.660735792873

RNNresults/X.Dataset3.RNN_3A_rmsprop

@@ -0,0 +1,12 @@
+RNN_3.2A Dataset3 Loss:binary_crossentropy Optimizer:rmsprop Dropout:No
+Epoch|Loss|Accuracy|Precision|Recall|Fscore
+1|4.11160773746|0.744738888889|0.985496704798|0.496788888889|0.660579600943
+2|4.11085833815|0.744911111111|0.986170541268|0.496788888889|0.660730910757
+3|4.11084675434|0.744916666667|0.98619229327|0.496788888889|0.660735792873
+4|4.11084652899|0.744916666667|0.98619229327|0.496788888889|0.660735792873
+5|4.11084649866|0.744916666667|0.98619229327|0.496788888889|0.660735792873
+6|4.11084649783|0.744916666667|0.98619229327|0.496788888889|0.660735792873
+7|4.1108464988|0.744916666667|0.98619229327|0.496788888889|0.660735792873
+8|4.11084649771|0.744916666667|0.98619229327|0.496788888889|0.660735792873
+9|4.11084649771|0.744916666667|0.98619229327|0.496788888889|0.660735792873
+10|4.11084649771|0.744916666667|0.98619229327|0.496788888889|0.660735792873

RNNresults/Z.Dataset3.RNN_1A_rmsprop

@@ -0,0 +1,12 @@
+RNN_1.2A Dataset3 Loss:binary_crossentropy Optimizer:rmsprop Dropout:No
+Epoch|Loss|Accuracy|Precision|Recall|Fscore
+1|0.407210527377|0.913938888889|0.985900796911|0.839888888889|0.907056416533
+2|1.32126423762|0.911833333333|0.986608901142|0.835|0.904495396281
+3|1.34785103439|0.912472222222|0.988100716587|0.835|0.905121796995
+4|1.35150305625|0.913994444444|0.991673374593|0.835|0.906617766813
+5|1.35529757347|0.914011111111|0.991712634274|0.835|0.906634173413
+6|1.34664266568|0.914305555556|0.991718426501|0.835588888889|0.90698361585
+7|1.34180664207|0.916511111111|0.991761557433|0.84|0.909594051544
+8|1.34144855539|0.916577777778|0.991762854145|0.840133333333|0.909672762271
+9|1.34134629022|0.916738888889|0.99176598618|0.840455555556|0.909862932958
+10|1.34128778595|0.916738888889|0.99176598618|0.840455555556|0.909862932958