In the previous column, we introduced the basic building blocks of an artificial neural network (ANN): neurons computing linear combinations of inputs, followed by a nonlinear activation function 1.