The second RL theory workshop (co-located with COLT 2024)
![Hiking photo](/hiking-pic.jpg)
Our second* RL theory workshop took place 27-29 June 2024 at the University of Alberta. Project Eva, Jack Mayo and Daniel Abrahamian’s startup, sponsored the event.
David Janz, Alex Ayoub & Csaba Szepesvári
*Recordings from the previous workshop are available here.
Talks list
June 28th am:
- Truncated Variance Reduced Value Iteration, 1hr, Ishani Aniruddha Karmarkar
- Optimistic Q-learning for average reward and episodic RL, 1hr, Shipra Agrawal
- Uncertainty-Aware Reward-Free Exploration with General Function Approximation, 30min, Dongruo Zhou
- On the Instance-dependent Sample Complexity of Tabular RL, 30min Kevin Jamieson
June 28th pm:
- Scalable Online Exploration via Coverability, 30min Philip Amortila, paper link
- Statistical and Algorithmic Reductions for Reinforcement Learning From Rich Observations, 30min, Philip Amortila
- Bisimulation Metrics are Optimal Transport Distances, and Can be Computed Efficiently, 45min Gergely Neu
- Towards Principled, Practical Policy Gradient for Bandits and Tabular MDPs, 30min Sharan Vaswani
- Self-Play Preference Optimization for Language Model Alignment, 30min Quanquan Gu
June 29th am:
- A Computationally Efficient Algorithm for Infinite-Horizon Average Reward Reinforcement Learning with Linear MDPs, 1hr, Ki Hong
- Offline Reinforcement Learning: Role of State Aggregation and Trajectory Data, 1hr, Zeyu Jia, paper link
- Efficient exploration in deep RL via utility theory, 45min, Brendan O’Donoghue
June 29th pm:
- Robust Online Learning in the Presence of Outliers, 30min, Tim van Erven, paper link
- Reinforcement Learning in Mean Field Games: the pitfalls and promises, 30min, Niao He
- When are Offline Multi-Agent Games Solvable?, 30min, Simon Du