Dataset

XOR Problem
Circle
Spiral
Clusters
Draw Custom

Network Architecture

Training

Network Architecture

Decision Boundary & Data

Training Stats

Epoch
0
Loss
0.000
Accuracy
0%
Speed
0 ep/s

About Neural Networks

Neural networks learn patterns through backpropagation:

  1. Forward pass computes predictions
  2. Loss measures error
  3. Backprop calculates gradients
  4. Weights update to reduce loss

The decision boundary shows what the network has learned. Blue = Class 0, Orange = Class 1.

Tips

XOR needs hidden layers
Circle shows non-linear separation
Spiral is the hardest problem
• More layers ≠ always better
• Learning rate affects convergence