Neural Networks from Scratch 🕸️

These notebooks explore Neural Networks, from the very basic Perceptron and its ability to classify linearly, to a DNN. All "from scratch", based on awesome tutorials from Guillaume Saint-Cirgue (Machine Learnia youtube channel) and more (see links below).

Notebook names :

1. Basic Perceptron

2. Neural Network with 1 hidden layer

3. Neural Network with 1 hidden layer and a multiclass output - softmax activation function

4. Neural Network with N layers

5. Neural Network with N layers and a multiclass output - softmax activation function

6. Neural Network with N layers as Class Objects, binary and multinomial classification

Python script :

DNN_from_scratch.py

Sources :

• for the mathematical notations and equations :

  • Machine Learnia's Youtube channel, deeplearning playlist

• for the multiclass classification : I based the DL_2layers_multiclass notebook on previous notebooks (DL_basics_1neuron & DL_2layers) and different sources from internet :

  • an example of multiclass DNN python code
  • Detailed equations & python code
  • Cross-Entropy calculation
  • Courses notes on deeplearning

The equations used for the N-layer Neural Network are defined above :