It's all about Origami
Neural networks explained by folding paper, a series in two parts.
Posts
It's all about the Entropy of Coffee
How entropy and thermodynamics explain the way neural networks classify (from a cup of scalding coffee to softmax).
It's all about Slinky
Backpropagation revisited through the physics of a spring toy (how perturbations propagate, dampen, or explode through a neural network).
It's all about Gas
Overfitting, underfitting, regularisation and the bias-variance tradeoff explained through the physics of gas.