Scipediacontent: Scipediacontent moved page Draft Content 849960942 to Krause et al 2019a

2021-02-02T06:35:29Z

Scipediacontent moved page Draft Content 849960942 to Krause et al 2019a

Scipediacontent: Created page with " == Abstract == The optimization of expensive to evaluate, black-box, mixed-variable functions, i.e. functions that have continuous and discrete inputs, is a difficult and ye..."

2021-02-02T06:35:26Z

Created page with " == Abstract == The optimization of expensive to evaluate, black-box, mixed-variable functions, i.e. functions that have continuous and discrete inputs, is a difficult and ye..."

New page

== Abstract ==

The optimization of expensive to evaluate, black-box, mixed-variable functions, i.e. functions that have continuous and discrete inputs, is a difficult and yet pervasive problem in science and engineering. In Bayesian optimization (BO), special cases of this problem that consider fully continuous or fully discrete domains have been widely studied. However, few methods exist for mixed-variable domains and none of them can handle discrete constraints that arise in many real-world applications. In this paper, we introduce MiVaBo, a novel BO algorithm for the efficient optimization of mixed-variable functions combining a linear surrogate model based on expressive feature representations with Thompson sampling. We propose an effective method to optimize its acquisition function, a challenging problem for mixed-variable domains, making MiVaBo the first BO method that can handle complex constraints over the discrete variables. Moreover, we provide the first convergence analysis of a mixed-variable BO algorithm. Finally, we show that MiVaBo is significantly more sample efficient than state-of-the-art mixed-variable BO algorithms on several hyperparameter tuning tasks, including the tuning of deep generative models.

Comment: IJCAI 2020 camera-ready; 17 pages, extended version with supplementary material

== Original document ==

The different versions of the original document can be found in:

* [http://arxiv.org/abs/1907.01329 http://arxiv.org/abs/1907.01329]

* [http://dx.doi.org/10.24963/ijcai.2020/365 http://dx.doi.org/10.24963/ijcai.2020/365]

* [http://dx.doi.org/10.3929/ethz-b-000385835 http://dx.doi.org/10.3929/ethz-b-000385835]

* [http://hdl.handle.net/20.500.11850/385835 http://hdl.handle.net/20.500.11850/385835]

* [https://www.ijcai.org/proceedings/2020/0365.pdf https://www.ijcai.org/proceedings/2020/0365.pdf]

* [https://dblp.uni-trier.de/db/conf/ijcai/ijcai2020.html#DaxbergerMT020 https://dblp.uni-trier.de/db/conf/ijcai/ijcai2020.html#DaxbergerMT020],
: [https://arxiv.org/abs/1907.01329 https://arxiv.org/abs/1907.01329],
: [https://arxiv.org/pdf/1907.01329.pdf https://arxiv.org/pdf/1907.01329.pdf],
: [https://www.ijcai.org/proceedings/2020/0365.pdf https://www.ijcai.org/proceedings/2020/0365.pdf],
: [https://ui.adsabs.harvard.edu/abs/2019arXiv190701329D/abstract https://ui.adsabs.harvard.edu/abs/2019arXiv190701329D/abstract],
: [https://www.ijcai.org/Proceedings/2020/0365 https://www.ijcai.org/Proceedings/2020/0365],
: [https://www.research-collection.ethz.ch/handle/20.500.11850/385835 https://www.research-collection.ethz.ch/handle/20.500.11850/385835],
: [https://export.arxiv.org/pdf/1907.01329 https://export.arxiv.org/pdf/1907.01329],
: [https://ru.arxiv.org/abs/1907.01329?context=stat.ML https://ru.arxiv.org/abs/1907.01329?context=stat.ML],
: [https://fr.arxiv.org/abs/1907.01329 https://fr.arxiv.org/abs/1907.01329],
: [https://uk.arxiv.org/abs/1907.01329?context=stat.ML https://uk.arxiv.org/abs/1907.01329?context=stat.ML],
: [https://tw.arxiv.org/pdf/1907.01329 https://tw.arxiv.org/pdf/1907.01329],
: [https://export.arxiv.org/abs/1907.01329 https://export.arxiv.org/abs/1907.01329],
: [https://au.arxiv.org/pdf/1907.01329 https://au.arxiv.org/pdf/1907.01329],
: [https://au.arxiv.org/abs/1907.01329?context=stat https://au.arxiv.org/abs/1907.01329?context=stat],
: [https://academic.microsoft.com/#/detail/2954731832 https://academic.microsoft.com/#/detail/2954731832]

DOIS: 10.3929/ethz-b-000385835 10.24963/ijcai.2020/365

← Older revision	Revision as of 06:35, 2 February 2021
(No difference)

Krause et al 2019a - Revision history

Scipediacontent: Scipediacontent moved page Draft Content 849960942 to Krause et al 2019a

Scipediacontent: Created page with " == Abstract == The optimization of expensive to evaluate, black-box, mixed-variable functions, i.e. functions that have continuous and discrete inputs, is a difficult and ye..."