Offres ares à Clermont-FerrandCréer des alertes d'emploi

A tiny example makes this concrete. Say you have the points (1, 3), (2, 5), (3, 7) and want a program that predicts y from x. The network is just `y = w·x + b`, with `w` and `b` starting at 0. The code says: • predict y for each x, • measure how far off the predictions are, • nudge `w` and `b` slightly in the direction that shrinks the error repeat