Do you think sigmoid.social should have LaTeX rendering in the interface like Mathstodon? Would be useful to some of us, right?
| Scholar | https://scholar.google.com/citations?user=5CqEn6YAAAAJ |
Ashish Gaurav
- 95 Followers
- 194 Following
- 38 Posts
#PhD student at #VectorInstitute / #UniversityOfWaterloo. (Studying #RL #ReinforcementLearning)
Tagging @thegradient (I think they are currently running this, but please correct me if I am wrong)
Hi everyone, I am a PhD-CS candidate at Univ. of Waterloo studying reinforcement learning and decision making.
More concretely, I currently work on inverse reinforcement learning to learn constraint functions. I am also interested in reward refinement strategies (eg. reward redistribution / credit assignment). I think that these are two ways to specify human preferences better. Better specification of human preferences can improve the adoption of RL as a practical procedure.