Self-tuning portfolio-based Bayesian optimization
作者：
Highlights：
• The work proposes a new portfolio based Bayesian Optimization method (SETUP-BO).
• SETUP-BO improves the performance of GP-Hedge while keeping its easiness to use.
• SETUP-BO uses a Thompson Sampling to autonomous tune all hyperparameters.
• The experiments showed that SETUP-BO is a valid alternative for various black box optimization tasks.
摘要
•The work proposes a new portfolio based Bayesian Optimization method (SETUP-BO).•SETUP-BO improves the performance of GP-Hedge while keeping its easiness to use.•SETUP-BO uses a Thompson Sampling to autonomous tune all hyperparameters.•The experiments showed that SETUP-BO is a valid alternative for various black box optimization tasks.
论文关键词：Bayesian optimization,Acquisition functions,Portfolio allocation,Thompson sampling
论文评审过程：Received 28 November 2020, Revised 21 July 2021, Accepted 30 August 2021, Available online 25 September 2021, Version of Record 19 October 2021.
论文官网地址：https://doi.org/10.1016/j.eswa.2021.115847
