Summary of Doubly-robust Off-policy Evaluation with Estimated Logging Policy, by Kyungbok Lee et al.
Doubly-Robust Off-Policy Evaluation with Estimated Logging Policy
by Kyungbok Lee, Myunghee Cho Paik
First submitted to arxiv on: 2 Apr 2024
Categories
- Main: Machine Learning (stat.ML)
- Secondary: Machine Learning (cs.LG)
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 The proposed doubly-robust off-policy evaluation estimator, DRUnknown, addresses situations where both the logging policy and value function are unknown for Markov decision processes. This novel approach initially estimates the logging policy and then minimizes the asymptotic variance while considering the estimation effect of the logging policy to estimate the value function model. When the logging policy model is correct, DRUnknown achieves the smallest asymptotic variance within existing estimators. In cases where both models are correct, DRUnknown reaches the semiparametric lower bound. Experimental results in contextual bandits and reinforcement learning demonstrate the performance of DRUnknown compared to existing methods. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary The paper introduces a new way to evaluate how well a policy works without knowing what the original policy was or how good it is. This is helpful when trying to decide which policy to use in situations where you don’t have all the information. The method, called DRUnknown, first tries to figure out what the original policy was and then uses that to estimate how good the new policy is. When done correctly, this method does a great job of estimating how well the policy works. |
Keywords
* Artificial intelligence * Reinforcement learning
