Figure 1. A feedforward network with two layers, with two nodes per layer, and three inputs.
W is the weight matrix between the hidden and output layer, while w is the weight matrix between the inputs
and the hidden layer. The network has two bias nodes with value 1.