🧠 Two-Spiral Problem: Backpropagation Training of an MLP in 3D
A trainable multilayer perceptron shown as a 3D network with a live decision floor, backpropagation pulses, loss curves and per-layer gradients on the two-spiral problem.
⚙ Under the hood
A small multilayer perceptron learns to separate two interleaved spirals with real forward passes, backpropagation and Adam or SGD, shown as a 3D network with a live decision surface, loss curves and gradient flow.
neural networkbackpropagationmultilayer perceptrondecision boundarygradient descentloss curvemachine learning
3D · Three.js / WebGL renderer · 60 FPS target · runs fully client-side, no install