Three pieces of math — derivatives, the chain rule, and log loss — quietly decide how every neural network learns from being wrong Continue reading on Towards AI »