C2: Backpropagation Walkthrough
Step through the Chain Rule to see how error gradients flow backward through the network.
Target Output:
1.0
1. Run Forward Pass
2. Start Backprop
◀ Prev
Next ▶
▶ Auto Play
Backprop Progress
0 / 0
Reset All
Ready. Click "Run Forward Pass" to begin.
Local Gradients
σ(z)
→
σ(1-σ)
y = w·x
→
∂y/∂w = x
y = Σxᵢ
→
∂y/∂xᵢ = 1
L = ½(t-y)²
→
∂L/∂y = y-t
Network State
Current Calculation
∂L/∂w = ...
...
Waiting to start...
Forward Signal (Activation)
Backward Gradient (Error)
Bias Terms
Weights