MENU

Fun & Interesting

Applying the Momentum Optimizer to Gradient Descent

John Lins 3,651 lượt xem 3 years ago
Video Not Working? Fix It Now

CORRECTION: we are SUBTRACTING the learning_rate * gradient of last iteration(s) NOT adding.

We are still adding the "tweaks" because, by definition, the tweaks are already the negative of the learning_rate * gradient of last iteration(s).


#machinelearning #ml

Comment