Summary of Thompson Sampling For Infinite-horizon Discounted Decision Processes, by Daniel Adelman et al.
Thompson Sampling for Infinite-Horizon Discounted Decision Processes
by Daniel Adelman, Cagla Keceli, Alba V. Olivares-Nadal
First submitted to arxiv on: 14 May 2024
Categories
- Main: Machine Learning (stat.ML)
- Secondary: Machine Learning (cs.LG); Optimization and Control (math.OC)
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 In this research paper, scientists study a Markov decision process with an unknown parameter, using Thompson sampling as the algorithm to investigate its asymptotic behavior. The standard definition of regret may not be suitable for evaluating policies in realistic settings, so they develop a new metric called expected residual regret that measures regret against the optimal reward moving forward from the current period. They show that the expected residual regret of Thompson sampling is upper bounded by a term that converges exponentially fast to 0 and provide conditions under which the posterior sampling error of Thompson sampling converges to 0 almost surely. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper explores how machines can learn from their actions in uncertain situations. The researchers design an algorithm called Thompson sampling, which helps make good decisions based on incomplete information. They also create a new way to measure the quality of these decisions, called expected residual regret. This metric helps them understand when the algorithm is making progress and when it’s stuck. By studying how well this algorithm works, the scientists provide a better understanding of how machines can learn from their experiences. |
