Bayesian Optimisation: Which Constraints Matter?
Xietao Wang Lin, Juan Ungredda, Max Butler, James Town, Alma Rahat, Hemant Singh, Juergen Branke · 2025
Read Bayesian Optimisation: Which Constraints Matter? by Xietao Wang Lin, Juan Ungredda, Max Butler, James Town, Alma Rahat, Hemant Singh, Juergen Branke (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2512-17569.txt.
Description
Xietao Wang Lin and coauthors work on Bayesian optimization for costly black box problems that carry several constraints. In many such problems the goal and each constraint can be tested on their own. Only a few constraints actually bind at the best solution, so testing all of them every time wastes effort.
The team builds new versions of the Knowledge Gradient rule that score each constraint on its own. The new rules sample only the constraints worth knowing about, and a faster variant pairs the idea with expected improvement. In benchmark tests the decoupled methods beat the coupled ones and the prior state of the art, PESC, with the gap widest when constraints differ in cost. The savings matter wherever each test burns time, money, or machine hours.
Catalog record
| CREATOR | Xietao Wang Lin, Juan Ungredda, Max Butler, James Town, Alma Rahat, Hemant Singh, Juergen Branke |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 42 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2512-17569