XOR Network Simulator

The XOR problem is the classic example of a function that cannot be computed by a single McCulloch-Pitts neuron. XOR outputs 1 when exactly one of its inputs is 1 — but the four input patterns are not linearly separable. However, a 2-layer network with 5 neurons solves it easily: two hidden neurons decompose XOR into sub-problems (x₁ AND NOT x₂ and NOT x₁ AND x₂), and the output neuron combines them with OR. Toggle the inputs below and watch the signals propagate through the network.