Neural Net Playground 3D

How to Play

1. Build and watch a neural network in 3D: neurons are spheres, connections are lines whose opacity reflects the weight.
2. Take on the challenges (e.g. XOR classifier): set the inputs and watch the forward pass propagate activations layer by layer.
3. The goal is to make the network produce the correct output for each of the challenge’s input vectors.
4. Rotate the scene to inspect weights and activations from every angle.