gradient descent

An optimization algorithm that minimizes error iteratively

“Gradient descent adjusts weights to reduce the loss function.”

Origin: From Latin gradus (step) + descendere (to go down), from de- (down) + scandere (to climb)