Summary of Federated Offline Policy Learning, by Aldo Gael Carranza et al.
Federated Offline Policy Learning
by Aldo Gael Carranza, Susan Athey
First submitted to arxiv on: 21 May 2023
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Distributed, Parallel, and Cluster Computing (cs.DC); Econometrics (econ.EM); Machine Learning (stat.ML)
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 approach addresses the challenge of learning personalized decision policies from observational bandit feedback data across multiple heterogeneous data sources. The method introduces a novel regret analysis that establishes finite-sample upper bounds on global and local regret for each data source. This is achieved by characterizing regret bounds through expressions of source heterogeneity and distribution shift. The paper also examines practical considerations in the federated setting, where a central server trains a policy without collecting raw data from individual sources. A policy learning algorithm is presented, which aggregates local policies trained using doubly robust offline policy evaluation strategies. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Imagine trying to learn how people make decisions based on different types of data. This paper figures out how to do that by combining information from many different places. It shows that this can be done in a way that makes sense and is reliable, even when the data is very different. The research also explores what happens when you’re trying to train a decision-making model without having all the original data. A new approach is developed for learning these decision policies, which combines information from each individual source. |
