Finding All {\epsilon}-Good Arms in Stochastic Bandits
Mason, Blake, Jain, Lalit, Tripathy, Ardhendu, Nowak, Robert
The pure-exploration problem in stochastic multi-armed bandits aims to find one or more arms with the largest (or near largest) means. Examples include finding an {\epsilon}-good arm, best-arm identification, top-k arm identification, and finding all arms with means above a specified threshold. However, the problem of finding all {\epsilon}-good arms has been overlooked in past work, although arguably this may be the most natural objective in many applications. For example, a virologist may conduct preliminary laboratory experiments on a large candidate set of treatments and move all {\epsilon}-good treatments into more expensive clinical trials. Since the ultimate clinical efficacy is uncertain, it is important to identify all {\epsilon}-good candidates. Mathematically, the all-{\epsilon}-good arm identification problem presents significant new challenges and surprises that do not arise in the pure-exploration objectives studied in the past. We introduce two algorithms to overcome these and demonstrate their great empirical performance on a large-scale crowd-sourced dataset of 2.2M ratings collected by the New Yorker Caption Contest as well as a dataset testing hundreds of possible cancer drugs.
Sep-11-2020
- Country:
- North America > United States
- New York (0.24)
- Wisconsin > Dane County
- Madison (0.14)
- North America > United States
- Genre:
- Research Report
- Experimental Study (0.33)
- New Finding (0.47)
- Research Report
- Industry:
- Technology: