Loading…
Pre-trained Gaussian Processes for Bayesian Optimization
Bayesian optimization (BO) has become a popular strategy for global optimization of expensive real-world functions. Contrary to a common expectation that BO is suited to optimizing black-box functions, it actually requires domain knowledge about those functions to deploy BO successfully. Such domain...
Saved in:
Published in: | arXiv.org 2024-08 |
---|---|
Main Authors: | , , , , , , , |
Format: | Article |
Language: | English |
Subjects: | |
Online Access: | Get full text |
Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
cited_by | |
---|---|
cites | |
container_end_page | |
container_issue | |
container_start_page | |
container_title | arXiv.org |
container_volume | |
creator | Wang, Zi Dahl, George E Swersky, Kevin Lee, Chansoo Nado, Zachary Gilmer, Justin Snoek, Jasper Ghahramani, Zoubin |
description | Bayesian optimization (BO) has become a popular strategy for global optimization of expensive real-world functions. Contrary to a common expectation that BO is suited to optimizing black-box functions, it actually requires domain knowledge about those functions to deploy BO successfully. Such domain knowledge often manifests in Gaussian process (GP) priors that specify initial beliefs on functions. However, even with expert knowledge, it is non-trivial to quantitatively define a prior. This is especially true for hyperparameter tuning problems on complex machine learning models, where landscapes of tuning objectives are often difficult to comprehend. We seek an alternative practice for setting these functional priors. In particular, we consider the scenario where we have data from similar functions that allow us to pre-train a tighter distribution a priori. We detail what pre-training entails for GPs using a KL divergence based loss function, and propose a new pre-training based BO framework named HyperBO. Theoretically, we show bounded posterior predictions and near-zero regrets for HyperBO without assuming the "ground truth" GP prior is known. To verify our approach in realistic setups, we collect a large multi-task hyperparameter tuning dataset by training tens of thousands of configurations of near-state-of-the-art deep learning models on popular image and text datasets, as well as a protein sequence dataset. Our results show that on average, HyperBO is able to locate good hyperparameters at least 3 times more efficiently than the best competing methods on both our new tuning dataset and existing multi-task BO benchmarks. |
format | article |
fullrecord | <record><control><sourceid>proquest</sourceid><recordid>TN_cdi_proquest_journals_2574547655</recordid><sourceformat>XML</sourceformat><sourcesystem>PC</sourcesystem><sourcerecordid>2574547655</sourcerecordid><originalsourceid>FETCH-proquest_journals_25745476553</originalsourceid><addsrcrecordid>eNpjYuA0MjY21LUwMTLiYOAtLs4yMDAwMjM3MjU15mSwCChK1S0pSszMS01RcE8sLS7OTMxTCCjKT04tLk4tVkjLL1JwSqxMBQv7F5Rk5mZWJZZk5ufxMLCmJeYUp_JCaW4GZTfXEGcP3YKi_MLS1OKS-Kz80qI8oFS8kam5iamJuRnQQuJUAQBd5zZo</addsrcrecordid><sourcetype>Aggregation Database</sourcetype><iscdi>true</iscdi><recordtype>article</recordtype><pqid>2574547655</pqid></control><display><type>article</type><title>Pre-trained Gaussian Processes for Bayesian Optimization</title><source>Publicly Available Content Database</source><creator>Wang, Zi ; Dahl, George E ; Swersky, Kevin ; Lee, Chansoo ; Nado, Zachary ; Gilmer, Justin ; Snoek, Jasper ; Ghahramani, Zoubin</creator><creatorcontrib>Wang, Zi ; Dahl, George E ; Swersky, Kevin ; Lee, Chansoo ; Nado, Zachary ; Gilmer, Justin ; Snoek, Jasper ; Ghahramani, Zoubin</creatorcontrib><description>Bayesian optimization (BO) has become a popular strategy for global optimization of expensive real-world functions. Contrary to a common expectation that BO is suited to optimizing black-box functions, it actually requires domain knowledge about those functions to deploy BO successfully. Such domain knowledge often manifests in Gaussian process (GP) priors that specify initial beliefs on functions. However, even with expert knowledge, it is non-trivial to quantitatively define a prior. This is especially true for hyperparameter tuning problems on complex machine learning models, where landscapes of tuning objectives are often difficult to comprehend. We seek an alternative practice for setting these functional priors. In particular, we consider the scenario where we have data from similar functions that allow us to pre-train a tighter distribution a priori. We detail what pre-training entails for GPs using a KL divergence based loss function, and propose a new pre-training based BO framework named HyperBO. Theoretically, we show bounded posterior predictions and near-zero regrets for HyperBO without assuming the "ground truth" GP prior is known. To verify our approach in realistic setups, we collect a large multi-task hyperparameter tuning dataset by training tens of thousands of configurations of near-state-of-the-art deep learning models on popular image and text datasets, as well as a protein sequence dataset. Our results show that on average, HyperBO is able to locate good hyperparameters at least 3 times more efficiently than the best competing methods on both our new tuning dataset and existing multi-task BO benchmarks.</description><identifier>EISSN: 2331-8422</identifier><language>eng</language><publisher>Ithaca: Cornell University Library, arXiv.org</publisher><subject>Artificial neural networks ; Bayesian analysis ; Datasets ; Decision theory ; Experimentation ; Iterative methods ; Mathematical models ; Neural networks ; Optimization ; Parameter sensitivity ; Training ; Tuning</subject><ispartof>arXiv.org, 2024-08</ispartof><rights>2024. This work is published under http://creativecommons.org/licenses/by/4.0/ (the “License”). Notwithstanding the ProQuest Terms and Conditions, you may use this content in accordance with the terms of the License.</rights><oa>free_for_read</oa><woscitedreferencessubscribed>false</woscitedreferencessubscribed></display><links><openurl>$$Topenurl_article</openurl><openurlfulltext>$$Topenurlfull_article</openurlfulltext><thumbnail>$$Tsyndetics_thumb_exl</thumbnail><linktohtml>$$Uhttps://www.proquest.com/docview/2574547655?pq-origsite=primo$$EHTML$$P50$$Gproquest$$Hfree_for_read</linktohtml><link.rule.ids>780,784,25753,37012,44590</link.rule.ids></links><search><creatorcontrib>Wang, Zi</creatorcontrib><creatorcontrib>Dahl, George E</creatorcontrib><creatorcontrib>Swersky, Kevin</creatorcontrib><creatorcontrib>Lee, Chansoo</creatorcontrib><creatorcontrib>Nado, Zachary</creatorcontrib><creatorcontrib>Gilmer, Justin</creatorcontrib><creatorcontrib>Snoek, Jasper</creatorcontrib><creatorcontrib>Ghahramani, Zoubin</creatorcontrib><title>Pre-trained Gaussian Processes for Bayesian Optimization</title><title>arXiv.org</title><description>Bayesian optimization (BO) has become a popular strategy for global optimization of expensive real-world functions. Contrary to a common expectation that BO is suited to optimizing black-box functions, it actually requires domain knowledge about those functions to deploy BO successfully. Such domain knowledge often manifests in Gaussian process (GP) priors that specify initial beliefs on functions. However, even with expert knowledge, it is non-trivial to quantitatively define a prior. This is especially true for hyperparameter tuning problems on complex machine learning models, where landscapes of tuning objectives are often difficult to comprehend. We seek an alternative practice for setting these functional priors. In particular, we consider the scenario where we have data from similar functions that allow us to pre-train a tighter distribution a priori. We detail what pre-training entails for GPs using a KL divergence based loss function, and propose a new pre-training based BO framework named HyperBO. Theoretically, we show bounded posterior predictions and near-zero regrets for HyperBO without assuming the "ground truth" GP prior is known. To verify our approach in realistic setups, we collect a large multi-task hyperparameter tuning dataset by training tens of thousands of configurations of near-state-of-the-art deep learning models on popular image and text datasets, as well as a protein sequence dataset. Our results show that on average, HyperBO is able to locate good hyperparameters at least 3 times more efficiently than the best competing methods on both our new tuning dataset and existing multi-task BO benchmarks.</description><subject>Artificial neural networks</subject><subject>Bayesian analysis</subject><subject>Datasets</subject><subject>Decision theory</subject><subject>Experimentation</subject><subject>Iterative methods</subject><subject>Mathematical models</subject><subject>Neural networks</subject><subject>Optimization</subject><subject>Parameter sensitivity</subject><subject>Training</subject><subject>Tuning</subject><issn>2331-8422</issn><fulltext>true</fulltext><rsrctype>article</rsrctype><creationdate>2024</creationdate><recordtype>article</recordtype><sourceid>PIMPY</sourceid><recordid>eNpjYuA0MjY21LUwMTLiYOAtLs4yMDAwMjM3MjU15mSwCChK1S0pSszMS01RcE8sLS7OTMxTCCjKT04tLk4tVkjLL1JwSqxMBQv7F5Rk5mZWJZZk5ufxMLCmJeYUp_JCaW4GZTfXEGcP3YKi_MLS1OKS-Kz80qI8oFS8kam5iamJuRnQQuJUAQBd5zZo</recordid><startdate>20240802</startdate><enddate>20240802</enddate><creator>Wang, Zi</creator><creator>Dahl, George E</creator><creator>Swersky, Kevin</creator><creator>Lee, Chansoo</creator><creator>Nado, Zachary</creator><creator>Gilmer, Justin</creator><creator>Snoek, Jasper</creator><creator>Ghahramani, Zoubin</creator><general>Cornell University Library, arXiv.org</general><scope>8FE</scope><scope>8FG</scope><scope>ABJCF</scope><scope>ABUWG</scope><scope>AFKRA</scope><scope>AZQEC</scope><scope>BENPR</scope><scope>BGLVJ</scope><scope>CCPQU</scope><scope>DWQXO</scope><scope>HCIFZ</scope><scope>L6V</scope><scope>M7S</scope><scope>PIMPY</scope><scope>PQEST</scope><scope>PQQKQ</scope><scope>PQUKI</scope><scope>PRINS</scope><scope>PTHSS</scope></search><sort><creationdate>20240802</creationdate><title>Pre-trained Gaussian Processes for Bayesian Optimization</title><author>Wang, Zi ; Dahl, George E ; Swersky, Kevin ; Lee, Chansoo ; Nado, Zachary ; Gilmer, Justin ; Snoek, Jasper ; Ghahramani, Zoubin</author></sort><facets><frbrtype>5</frbrtype><frbrgroupid>cdi_FETCH-proquest_journals_25745476553</frbrgroupid><rsrctype>articles</rsrctype><prefilter>articles</prefilter><language>eng</language><creationdate>2024</creationdate><topic>Artificial neural networks</topic><topic>Bayesian analysis</topic><topic>Datasets</topic><topic>Decision theory</topic><topic>Experimentation</topic><topic>Iterative methods</topic><topic>Mathematical models</topic><topic>Neural networks</topic><topic>Optimization</topic><topic>Parameter sensitivity</topic><topic>Training</topic><topic>Tuning</topic><toplevel>online_resources</toplevel><creatorcontrib>Wang, Zi</creatorcontrib><creatorcontrib>Dahl, George E</creatorcontrib><creatorcontrib>Swersky, Kevin</creatorcontrib><creatorcontrib>Lee, Chansoo</creatorcontrib><creatorcontrib>Nado, Zachary</creatorcontrib><creatorcontrib>Gilmer, Justin</creatorcontrib><creatorcontrib>Snoek, Jasper</creatorcontrib><creatorcontrib>Ghahramani, Zoubin</creatorcontrib><collection>ProQuest SciTech Collection</collection><collection>ProQuest Technology Collection</collection><collection>Materials Science & Engineering Collection</collection><collection>ProQuest Central (Alumni Edition)</collection><collection>ProQuest Central UK/Ireland</collection><collection>ProQuest Central Essentials</collection><collection>AUTh Library subscriptions: ProQuest Central</collection><collection>Technology Collection</collection><collection>ProQuest One Community College</collection><collection>ProQuest Central Korea</collection><collection>SciTech Premium Collection</collection><collection>ProQuest Engineering Collection</collection><collection>Engineering Database</collection><collection>Publicly Available Content Database</collection><collection>ProQuest One Academic Eastern Edition (DO NOT USE)</collection><collection>ProQuest One Academic</collection><collection>ProQuest One Academic UKI Edition</collection><collection>ProQuest Central China</collection><collection>Engineering Collection</collection></facets><delivery><delcategory>Remote Search Resource</delcategory><fulltext>fulltext</fulltext></delivery><addata><au>Wang, Zi</au><au>Dahl, George E</au><au>Swersky, Kevin</au><au>Lee, Chansoo</au><au>Nado, Zachary</au><au>Gilmer, Justin</au><au>Snoek, Jasper</au><au>Ghahramani, Zoubin</au><format>book</format><genre>document</genre><ristype>GEN</ristype><atitle>Pre-trained Gaussian Processes for Bayesian Optimization</atitle><jtitle>arXiv.org</jtitle><date>2024-08-02</date><risdate>2024</risdate><eissn>2331-8422</eissn><abstract>Bayesian optimization (BO) has become a popular strategy for global optimization of expensive real-world functions. Contrary to a common expectation that BO is suited to optimizing black-box functions, it actually requires domain knowledge about those functions to deploy BO successfully. Such domain knowledge often manifests in Gaussian process (GP) priors that specify initial beliefs on functions. However, even with expert knowledge, it is non-trivial to quantitatively define a prior. This is especially true for hyperparameter tuning problems on complex machine learning models, where landscapes of tuning objectives are often difficult to comprehend. We seek an alternative practice for setting these functional priors. In particular, we consider the scenario where we have data from similar functions that allow us to pre-train a tighter distribution a priori. We detail what pre-training entails for GPs using a KL divergence based loss function, and propose a new pre-training based BO framework named HyperBO. Theoretically, we show bounded posterior predictions and near-zero regrets for HyperBO without assuming the "ground truth" GP prior is known. To verify our approach in realistic setups, we collect a large multi-task hyperparameter tuning dataset by training tens of thousands of configurations of near-state-of-the-art deep learning models on popular image and text datasets, as well as a protein sequence dataset. Our results show that on average, HyperBO is able to locate good hyperparameters at least 3 times more efficiently than the best competing methods on both our new tuning dataset and existing multi-task BO benchmarks.</abstract><cop>Ithaca</cop><pub>Cornell University Library, arXiv.org</pub><oa>free_for_read</oa></addata></record> |
fulltext | fulltext |
identifier | EISSN: 2331-8422 |
ispartof | arXiv.org, 2024-08 |
issn | 2331-8422 |
language | eng |
recordid | cdi_proquest_journals_2574547655 |
source | Publicly Available Content Database |
subjects | Artificial neural networks Bayesian analysis Datasets Decision theory Experimentation Iterative methods Mathematical models Neural networks Optimization Parameter sensitivity Training Tuning |
title | Pre-trained Gaussian Processes for Bayesian Optimization |
url | http://sfxeu10.hosted.exlibrisgroup.com/loughborough?ctx_ver=Z39.88-2004&ctx_enc=info:ofi/enc:UTF-8&ctx_tim=2024-12-25T15%3A49%3A04IST&url_ver=Z39.88-2004&url_ctx_fmt=infofi/fmt:kev:mtx:ctx&rfr_id=info:sid/primo.exlibrisgroup.com:primo3-Article-proquest&rft_val_fmt=info:ofi/fmt:kev:mtx:book&rft.genre=document&rft.atitle=Pre-trained%20Gaussian%20Processes%20for%20Bayesian%20Optimization&rft.jtitle=arXiv.org&rft.au=Wang,%20Zi&rft.date=2024-08-02&rft.eissn=2331-8422&rft_id=info:doi/&rft_dat=%3Cproquest%3E2574547655%3C/proquest%3E%3Cgrp_id%3Ecdi_FETCH-proquest_journals_25745476553%3C/grp_id%3E%3Coa%3E%3C/oa%3E%3Curl%3E%3C/url%3E&rft_id=info:oai/&rft_pqid=2574547655&rft_id=info:pmid/&rfr_iscdi=true |