1 · One perceptron. Pick a problem:
w₁
w₂
b
Train with Rosenblatt's rule
Reset to 0
2 · XOR with a hidden layer: 2 inputs → 2 hidden perceptrons → 1 output, trained by backpropagation
Train from a random start
Colour = the network's output (orange near 1, blue near 0). Dashed lines = where each hidden perceptron switches.