Perceptron Playground

XOR · linear limits · backpropagation

Lesson path

  1. Train single neuron — watch it fail on XOR
  2. Train multi-layer network — solve XOR
  3. Step through one backprop update

You are watching a single "decision maker" try to draw one straight line that separates orange from purple dots. XOR needs opposite corners to match — no single line can do that. The line keeps shifting but never solves the puzzle.

ŷ=0 ✓ŷ=0 → 1 ✗ŷ=0 → 1 ✗ŷ=0 ✓

Purple = 0 · Amber = 1 · Background shows network confidence

Epoch 0

Press Train — the chart builds epoch by epoch.