🔄 Backpropagation Visualizer

Watch data flow forward, then gradients flow backward

① Select an Input (XOR Problem)

[0, 0]
Expected: 0
[0, 1]
Expected: 1
[1, 0]
Expected: 1
[1, 1]
Expected: 0

② Current State

?
Network Output
0
Expected
?
Loss (Error²)
0
Training Steps

③ Step Through the Process

1
→
2
→
3
→
4
1: Forward 2: Calculate Loss 3: Backward 4: Update Weights

④ The Network

Forward pass (data flows right)
Backward pass (gradients flow left)

⑤ What's Happening

Click "Forward Pass" to start
We'll walk through how the network processes an input, calculates error, and then figures out how to adjust each weight to reduce that error.

⑥ Gradients (How Much to Adjust Each Weight)

Positive gradient → weight is too high → decrease it
Negative gradient → weight is too low → increase it