search

facebookresearch / Pearl

A Production-ready Reinforcement Learning AI Agent Library brought by the Applied Reinforcement Learning team at Meta.
MIT License
2.58k stars 154 forks source link

Functionality for Stein variational policy gradient and/or regularization through the use of prior policies #44

Open leonhalgryn opened 8 months ago

leonhalgryn commented 8 months ago

Are there any plans to add functionality to allow using prior policies for regularization similar to that of the Stein variational policy gradient (SVPG) (SVPG paper available at: https://arxiv.org/abs/1704.02399)

GreatArcStudios commented 8 months ago

Yeah this would be pretty great if added. Looks like as of now they'd have to add it on a per algorithm basis as it is a modification to the loss function. Perhaps they could abstract away the loss into a module.

rodrigodesalvobraz commented 6 months ago

Thank you. This would be good but we are currently working on higher-priority items. We will leave the issue open and update it when we get to it. Thank you.