Summary of Multi-armed Bandits with Interference, by Su Jia et al.
Multi-Armed Bandits with Interference
by Su Jia, Peter Frazier, Nathan Kallus
First submitted to arxiv on: 2 Feb 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: 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 This paper tackles the issue of experimentation with interference in online platforms. The authors focus on the cumulative performance, which has been less well understood compared to the final output of a policy. They introduce the problem of Multi-armed Bandits with Interference (MABI), where the learner assigns an arm to each experimental unit over a time horizon. The reward of each unit depends on the treatments of all units, and the influence decays based on spatial distance. The authors also consider a setup where the reward functions are chosen by an adversary. They show that switchback policies achieve optimal expected regret against the best fixed-arm policy but have high variance due to not accounting for the number of units (N). To address this, they propose a cluster randomization policy whose regret is optimal in expectation and admits a high probability bound that vanishes as N grows. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper helps us understand how to run experiments online without interfering with each other. It’s like trying to find the best way to give people different treatments when we don’t know which one will work best. The problem is that the results of our experiment depend on what happens to all the people involved, not just one person. This makes it hard to predict the outcome and choose the best treatment. The authors came up with a new approach called cluster randomization that helps us make better decisions and avoid mistakes. |
Keywords
* Artificial intelligence * Probability
