Training a neural network is the process of determining the values of the constants -- called weights and biases -- that essentially define the network. You do this ...