search

ikostrikov / pytorch-a3c

PyTorch implementation of Asynchronous Advantage Actor Critic (A3C) from "Asynchronous Methods for Deep Reinforcement Learning".
MIT License
1.23k stars 279 forks source link

Have you tried out Shared YellowFin? #23

Closed ethancaballero closed 7 years ago

ethancaballero commented 7 years ago

Did it have significant effect compared to Shared Adam/RMSprop?

ikostrikov commented 7 years ago

Yeah. It didn't work for me. Probably, updates are too noisy. It might work better with A2C but I was never able to make A2C as fast as A3C.