← Back to Gallery

Backpropagation

Epoch: 0
Loss: -
Pattern: XOR
Neural Network Training

• Forward pass: compute activations
• Backward pass: compute gradients
• Update: adjust weights by gradients

Positive
Negative
Gradient flow