arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 1602.01064stat.MLcs.ITcs.LGcs.ROmath.IT
收藏

单任务和多任务优化的最小遗憾搜索

Minimum Regret Search for Single- and Multi-Task Optimization

Universität Bremen(不来梅大学);Robert Bosch GmbH(罗伯特·博世有限公司)

Jan Hendrik Metzen

摘要:

我们提出了最小遗憾搜索(MRS),一种新的贝叶斯优化获取函数。MRS与熵搜索(ES)等信息理论方法有相似之处。然而,虽然ES的目的是在每个查询在最大限度地提高信息增益方面的全球最大值,MRS的目的是在最小化的预期简单的遗憾,其最终建议的最佳。虽然经验ES和MRS在大多数情况下执行类似,但MRS比ES产生更少的具有高简单遗憾的离群值。我们提供了一个合成的单任务优化问题,以及一个模拟的多任务机器人控制问题的实证结果。

英文摘要:

We propose minimum regret search (MRS), a novel acquisition function for Bayesian optimization. MRS bears similarities with information-theoretic approaches such as entropy search (ES). However, while ES aims in each query at maximizing the information gain with respect to the global maximum, MRS aims at minimizing the expected simple regret of its ultimate recommendation for the optimum. While empirically ES and MRS perform similar in most of the cases, MRS produces fewer outliers with high simple regret than ES. We provide empirical results both for a synthetic single-task optimization problem as well as for a simulated multi-task robotic control problem.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑