How a network computes, why gradients vanish, and what makes depth trainable
Everything so far, once, on real numbers. A batch goes in, two hidden layers and a head act on it, and predictions come out. This is the whole of what a trained model does when it is used. No gradient exists yet, nothing is being learned, and the loss at the end is only a number you read.
🔒 This is a Pro lesson — the interactive figure, worked examples, quiz and practice open with Pro access.
▶ The Forward Pass, End to End