ml-math

This is not a Machine Learning framework.

This is an implementation of fundamental mathematics of ML (specifically, Neural Networks) just for fun.

Run

$ python main.py

Contents

• main.py - training experiment: architecture and hyperparameters
• activation.py - activation functions and their derivatives
• algebra.py - vector and matrix operations
• data.csv - Iris dataset for training/evaluation
• data.py - data loading, shuffling, normalisation
• metrics.py - evaluation metrics and loss functions
• propagation.py - backpropagation algorithm
• train.py - parameter initialisation and model training

To-do (maybe)

• Train/evaluation data split
• Stochastic gradient descent
• More activation functions
• Softmax
• Regularisation
• Refactor all the things
• Optimise all the things