Building Neural Networks From Scratch
Playing around with neural networks!
These two Jupyter notebooks implement neural networks from scratch without Tensorflow or other machine learning packages. This is significant beacuse there are many neat tools out there that are typically used to build neural networks, but its inner processes are hidden to developers and the concepts underpinning neural networks can sometimes be forgotten.
By manually implementing a neural network, I must go through the rather laborious (but fruitful!) task of creating classes of Neurons, Layers, and Neural Networks, and implementing my own activation functions and back-propogation algorithms with stochastic gradient descent.
What it Actually Does
Trains two separate neural networks to 'learn' the XOR problem and multiplication.