Code

Modules

• neural_network module
• exercise module
• activation_and_output_functions module
• gradient_descent module
• mnist_tools module

Code structure

The code is organised as follows:

• The core implementation of the neural network is found in neural_network module;
• The exercise module is a demonstration of the neural_network module module using the MNIST data set;
• The activation_and_output_functions module and gradient_descent module provide elemental functions that support the neural_network module. However, if any function in these two modules are needed, it suffices to import neural_network module, because it already imports all functions from these modules;
• The mnist_tools module provide functions specific to the MNIST data set. It is essential to the exercise module, but it also provides a function that displays a single character encoded in the MNIST data set.