PDE-Net icon indicating copy to clipboard operation
PDE-Net copied to clipboard

Optimization Method

Open ankitesh97 opened this issue 4 years ago • 1 comments

I was wondering why did you use BFGS optimization instead of inbuilt ADAM/Gradient descent optimization method in pytorch?

ankitesh97 avatar Jun 01 '20 22:06 ankitesh97

I was wondering why did you use BFGS optimization instead of inbuilt ADAM/Gradient descent optimization method in pytorch?

BFGS with line search converges quicker. ADAM and SGD with a fixed learning rate are not stable.

ZichaoLong avatar Jun 12 '20 04:06 ZichaoLong