Stable Softmax & Log-Sum-Exp

Multivariate calculus from first principles

A classifier's final layer produces a vector of raw scores called logits, one per class. Softmax turns those logits into probabilities: numbers that are never negative and always add up to 1. The class with the largest logit gets the largest share of the probability, and because the formula is built from exponentials, even a small gap between logits can turn into a lopsided split.

🔒 This is a Pro lesson — the interactive figure, worked examples, quiz and practice open with Pro access.