Python-Based Neural Network (PyBaNN) is my attempt to implement several neural networks encountered in Neural Networks and Deep Learning in Python as I worked through the chapters.

• Python 3
• NumPy
• SciPy

To run, simply type:

python3 pybann.py protocol hidden_layer epoch batch_size

where

• protocol is the neural network protocol you wish to run
• hidden_layer is a list of integers, specifying the number of nodes within each hidden layer (i.e. layers besides the input and output layers)
• epoch is the number of times to repeat the training
• During training, we break the training data into small batches and train on each of them in turn. batch_size is the size of these batches

The available neural network protocol which has been implemented at current are in the network folder. For details on each protocol, see Neural Networks and Deep Learning.

You can input your own protocol. Implement your protocol as a Network object, with train and evaluate functions, then place your .py file in the network folder. You can then call your protocol by passing the name (excluding the .py extension!) of your protocol into PyBaNN.

The module for implemented protocols are kept in the network folder.

• Feedforward neural network

All datasets can be found in the data folder, and all have the same structure.

• Handwriting images: This dataset is from the MNIST database, adapted by Michael Nielson, the author of Neural Networks and Deep Learning. You can also get it directly here.
• Fashion article images: This dataset is from Zalando's article images, adapted by me to have the same format.