Backpropagation

The chain rule at scale

An interactive lesson with 6 steps, about 5 minutes: you predict, drag and solve, with instant feedback on every answer. It runs in your browser and needs JavaScript.

What you’ll learn

• Backpropagation uses the chain rule to find every weight’s gradient.
• For $L = (wx - t)^2$: $\partial L/\partial w = 2(wx - t)x$.
• Each update moves weights a small step against the gradient.

Steps

1. Blame flows backwards
2. Compute a gradient
3. Through two layers
4. Vanishing gradients
5. Choosing the step size
6. One update