How a network computes, why gradients vanish, and what makes depth trainable
Everything this course has built so far is static. A forward pass turns a batch of inputs into predictions, a loss measures how wrong they are, and the backward pass of dl.15 turns that one number into a gradient for every parameter in the model. Nothing has moved yet.
🔒 This is a Pro lesson — the interactive figure, worked examples, quiz and practice open with Pro access.
▶ The Training Loop, One Step at a Time