modern-fortran / neural-fortran

A parallel framework for deep learning
MIT License
395 stars 82 forks source link

Implement RMSprop in example/quadratic.f90 #136

Closed milancurcic closed 1 year ago

milancurcic commented 1 year ago

Let's make the first implementation in the example program and then we'll have a better idea how we need to refactor the library to integrate it there.

Reference

Spnetic-5 commented 1 year ago

Hello @milancurcic, As you know RMSProp invloves retrieving the weights and gradients of the network's layers, in order to update the weights based on the gradients. I don't see implementation of these two in network code, so how are we planning the implemention exactly.

milancurcic commented 1 year ago

Yes, see #140 where I added the example program. In a nutshell, we need to iterate over each layer and use select type to access the internal data structure. There are a few places in the library where the same pattern is used.

milancurcic commented 1 year ago

Done by #144