Summary of Preferential Multi-objective Bayesian Optimization, by Raul Astudillo et al.
Preferential Multi-Objective Bayesian Optimization
by Raul Astudillo, Kejun Li, Maegan Tucker, Chu Xin Cheng, Aaron D. Ames, Yisong Yue
First submitted to arxiv on: 20 Jun 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Optimization and Control (math.OC); 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 Preferential Bayesian optimization (PBO) is a framework designed to optimize decision-makers’ latent preferences over available design choices. Unlike existing work in PBO that assumes single-objective functions, our proposed framework addresses the gap by handling multiple objectives. We introduce dueling scalarized Thompson sampling (DSTS), a multi-objective extension of the popular dueling Thompson algorithm, which can be applied beyond PBO settings. DSTS outperforms benchmarks across four synthetic test functions and two simulated tasks in exoskeleton personalization and driving policy design, demonstrating its effectiveness. Furthermore, we prove that DSTS is asymptotically consistent, providing a first-ever convergence guarantee for dueling Thompson sampling in the PBO setting. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper introduces a new way to help decision-makers make better choices by considering multiple things they want at the same time. Usually, people try to optimize one thing while ignoring others, but this framework allows us to balance different goals. We developed an algorithm called dueling scalarized Thompson sampling that does this balancing act well and tested it on several tasks related to robotic devices and autonomous driving policy design. The results show that our method performs better than existing approaches in these scenarios. This is important because it provides a way to ensure that the choices we make are consistent with what we really want. |
Keywords
* Artificial intelligence * Optimization
