Neural Network Simulator (3D)

This is the full 3D companion to the 2D neural-network simulator: the same tiny feedforward network — 2 inputs, a configurable hidden layer, 1 output — learning the classic XOR problem through sigmoid activations and per-weight gradient backpropagation, exactly as in the original. What's new is the third dimension: instead of a flat 2D scatter of decision points, the network's output is plotted as real height over the entire input plane, so the XOR decision boundary becomes a surface you can orbit around and watch physically rise and fold as training progresses. The network diagram above the surface shows every weight as a coloured, live-updating line — blue for positive, red for negative — so you can see the same numbers driving both views at once.