Bayesian optimization with a finite budget: An approximate dynamic programming approach
We consider the problem of optimizing an expensive objective function when a finite budget of total evaluations is prescribed. In that context, the optimal solution strategy for Bayesian optimization can be formulated as a dynamic programming instance. This results in a complex problem with uncounta...
| Prif Awduron: | , , |
|---|---|
| Awduron Eraill: | |
| Fformat: | Erthygl |
| Cyhoeddwyd: |
Neural Information Processing Systems Foundation
2018
|
| Mynediad Ar-lein: | http://hdl.handle.net/1721.1/115164 https://orcid.org/0000-0003-4222-5358 https://orcid.org/0000-0003-2156-9338 |
| _version_ | 1826209124296163328 |
|---|---|
| author | Wolpert, David H. Lam, Remi Roger Alain Paul Willcox, Karen E |
| author2 | Massachusetts Institute of Technology. Department of Aeronautics and Astronautics |
| author_facet | Massachusetts Institute of Technology. Department of Aeronautics and Astronautics Wolpert, David H. Lam, Remi Roger Alain Paul Willcox, Karen E |
| author_sort | Wolpert, David H. |
| collection | MIT |
| description | We consider the problem of optimizing an expensive objective function when a finite budget of total evaluations is prescribed. In that context, the optimal solution strategy for Bayesian optimization can be formulated as a dynamic programming instance. This results in a complex problem with uncountable, dimension-increasing state space and an uncountable control space. We show how to approximate the solution of this dynamic programming problem using rollout, and propose rollout heuristics specifically designed for the Bayesian optimization setting. We present numerical experiments showing that the resulting algorithm for optimization with a finite budget outperforms several popular Bayesian optimization algorithms. |
| first_indexed | 2024-09-23T14:17:37Z |
| format | Article |
| id | mit-1721.1/115164 |
| institution | Massachusetts Institute of Technology |
| last_indexed | 2024-09-23T14:17:37Z |
| publishDate | 2018 |
| publisher | Neural Information Processing Systems Foundation |
| record_format | dspace |
| spelling | mit-1721.1/1151642022-09-28T19:48:44Z Bayesian optimization with a finite budget: An approximate dynamic programming approach Wolpert, David H. Lam, Remi Roger Alain Paul Willcox, Karen E Massachusetts Institute of Technology. Department of Aeronautics and Astronautics Lam, Remi Willcox, Karen E We consider the problem of optimizing an expensive objective function when a finite budget of total evaluations is prescribed. In that context, the optimal solution strategy for Bayesian optimization can be formulated as a dynamic programming instance. This results in a complex problem with uncountable, dimension-increasing state space and an uncountable control space. We show how to approximate the solution of this dynamic programming problem using rollout, and propose rollout heuristics specifically designed for the Bayesian optimization setting. We present numerical experiments showing that the resulting algorithm for optimization with a finite budget outperforms several popular Bayesian optimization algorithms. 2018-05-02T15:29:40Z 2018-05-02T15:29:40Z 2016-12 2018-04-17T17:25:15Z Article http://purl.org/eprint/type/ConferencePaper http://hdl.handle.net/1721.1/115164 Lam, Rem, Karen Willcox, and David H. Wolpert. "Bayesian Optimization with a Finite Budget: An Approximate Dynamic Programming Approach." Advances in Neural Information Processing Systems 29 (NIPS 2016), 5-12 December, 2016, Barcelona, Spain, Neural Information Processing Systems Foundation, 2016. © 2016 NIPS Foundation https://orcid.org/0000-0003-4222-5358 https://orcid.org/0000-0003-2156-9338 https://papers.nips.cc/paper/6188-bayesian-optimization-with-a-finite-budget-an-approximate-dynamic-programming-approach Advances in Neural Information Processing Systems 29 (NIPS 2016) Article is made available in accordance with the publisher's policy and may be subject to US copyright law. Please refer to the publisher's site for terms of use. application/pdf Neural Information Processing Systems Foundation Neural Information Processing Systems (NIPS) |
| spellingShingle | Wolpert, David H. Lam, Remi Roger Alain Paul Willcox, Karen E Bayesian optimization with a finite budget: An approximate dynamic programming approach |
| title | Bayesian optimization with a finite budget: An approximate dynamic programming approach |
| title_full | Bayesian optimization with a finite budget: An approximate dynamic programming approach |
| title_fullStr | Bayesian optimization with a finite budget: An approximate dynamic programming approach |
| title_full_unstemmed | Bayesian optimization with a finite budget: An approximate dynamic programming approach |
| title_short | Bayesian optimization with a finite budget: An approximate dynamic programming approach |
| title_sort | bayesian optimization with a finite budget an approximate dynamic programming approach |
| url | http://hdl.handle.net/1721.1/115164 https://orcid.org/0000-0003-4222-5358 https://orcid.org/0000-0003-2156-9338 |
| work_keys_str_mv | AT wolpertdavidh bayesianoptimizationwithafinitebudgetanapproximatedynamicprogrammingapproach AT lamremirogeralainpaul bayesianoptimizationwithafinitebudgetanapproximatedynamicprogrammingapproach AT willcoxkarene bayesianoptimizationwithafinitebudgetanapproximatedynamicprogrammingapproach |