Here is an introduction to gradient descent with back propagation, for Ruby, based on Andrej Karpathy's micrograd: https://github.com/rickhull/backprop