Neural Network Playground

Place points. Train a network. Watch it learn.

Interactive companion to Issue 6: Machines That Learn

← Back to Issue 6
Epoch: 0 Loss: -- Accuracy: --
Data & Decision Boundary
Click canvas to add points
Network Architecture
Too high = overshoots. Too low = learns slowly.
More neurons = more complex boundaries.
The 'squish function' that gives the network its power.
Training Loss
The network learned to classify! It discovered the pattern on its own.