Summary of Near-optimal Regret in Linear Mdps with Aggregate Bandit Feedback, by Asaf Cassel and Haipeng Luo and Aviv Rosenberg and Dmitry Sotnikov
Near-Optimal Regret in Linear MDPs with Aggregate Bandit Feedback
by Asaf Cassel, Haipeng Luo, Aviv Rosenberg, Dmitry Sotnikov
First submitted to arxiv on: 13 May 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: None
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary This paper explores Reinforcement Learning (RL) in scenarios where receiving individual rewards is impractical. The authors focus on RL with Aggregate Bandit Feedback (RL-ABF), which provides feedback at the end of an episode, rather than after each action. Building upon prior work in tabular settings, this study extends RL-ABF to linear function approximation and proposes two efficient algorithms: a value-based optimistic algorithm utilizing a Q-functions ensemble and randomization technique, and a policy optimization algorithm employing a novel hedging scheme over the ensemble. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper is about teaching machines to make good decisions without getting individual rewards for each step. Instead, they get feedback at the end of an episode. The researchers are working on a way to use this method in situations where we can’t give rewards after every action. They developed two new ways to do this: one that uses many different versions of Q-functions and randomizes its choices, and another that optimizes policies using a special kind of averaging. |
Keywords
» Artificial intelligence » Optimization » Reinforcement learning
