Teacher lihongyi from National Taiwan University - grade Descent 2

Gradient Descent

Learning Rate

We can do it in the actual process , Draw the picture on the right , Every time the parameter is updated ,loss The change of , Then choose the right one learning rate.

Stochastic Gradient Descent

Feature Scaling

Gradient Descent Theory

Taylor unfolds here to review 5328









It's like an empire age game , We can only find the local minimum . Never know the global minimum . Unless it's a convex function , The local minimum is the global minimum .