Training

Coming soon: backpropagation from first principles, optimizers, and why training is stable (or not).