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:
- Forward pass computes predictions
- Loss measures error
- Backprop calculates gradients
- 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
• Circle shows non-linear separation
• Spiral is the hardest problem
• More layers ≠ always better
• Learning rate affects convergence