Gradient descent

Applies to: python, general

Gradient descent updates parameters opposite the gradient to reduce a loss function.

w -= learning_rate * grad_w

See also: gradient, learning-rate