Summary of Contextual Linear Optimization with Bandit Feedback, by Yichun Hu et al.
Contextual Linear Optimization with Bandit Feedback
by Yichun Hu, Nathan Kallus, Xiaojie Mao, Yanchen Wu
First submitted to arxiv on: 26 May 2024
Categories
- Main: Machine Learning (stat.ML)
- Secondary: Machine Learning (cs.LG); Methodology (stat.ME)
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 paper, researchers tackle the problem of contextual linear optimization (CLO) in scenarios where only historical decision outcomes are available, rather than fully observed cost coefficient vectors. They propose induced empirical risk minimization (IERM), an offline learning algorithm that fits a predictive model to optimize downstream policy performance. The authors demonstrate fast-rate regret bounds for IERM with misspecified model classes and flexible optimization estimates. A byproduct of the theory is also a fast-rate regret bound for IERM with full feedback and misspecified policy class. Numerical experiments using a stochastic shortest path example show that different modeling choices can have varying effects, providing practical insights. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper helps solve a problem where we can only see the outcome of past decisions, not all the details behind them. The researchers create an algorithm to make better predictions by looking at what happened in the past and using that to make future decisions. They show how this algorithm works well even when the models used are not perfect. This could be useful for things like predicting traffic flow or optimizing routes. |
Keywords
» Artificial intelligence » Optimization
