search

rlworkgroup / garage

A toolkit for reproducible reinforcement learning research.
MIT License
1.85k stars 309 forks source link

Add HER to DQN #210

Open ryanjulian opened 6 years ago

ryanjulian commented 6 years ago

https://arxiv.org/abs/1707.01495

ryanjulian commented 6 years ago

@CatherineSue will this be supported by just using the HerReplayBuffer, or do we need append the DQN also as well?

CatherineSue commented 6 years ago

We need to append DQN. It should be simple since we already have the primitives. We don't have argmax policy but it's very easy to implement, too.

ryanjulian commented 4 years ago

@maliesa96 does #1338 imply this issue will be fixed, as long as DQN uses PathBuffer?

maliesa96 commented 4 years ago

Not out of the box, DQN will need some tweaking but probably nothing major.