最大值熵搜索的有效贝叶斯优化

题目：Max-value Entropy Search for EfficientBayesian Optimization

中文摘要：整篇文章主要是对熵搜索的技术性改造。

1. 采取先验分布的高斯过程假定

2. 针对熵搜索的评价函数（黑箱函数）进行最大值熵函数改造

3. 进一步通过一系列数理假定和采样方法的设定进行搜索方法的收敛性证明

4. 进行实际数据实验进一步验证有效性

   
   

英文摘要：Entropy Search (ES) and PredictiveEntropy Search (PES) are popular and empirically successful BayesianOptimization techniques. Both rely on a compelling information-theoreticmotivation, and maximize the information gained about the arg max of theunknown function; yet, both are plagued by the expensive computation forestimating entropies. We propose a new criterion, Max-value Entropy Search(MES), that instead uses the information about the maximum function value. Weshow relations of MES to other Bayesian optimization methods, and establish aregret bound. We observe that MES maintains or improves the good empiricalperformance of ES/PES, while tremendously lightening the computational burden.In particular, MES is much more robust to the number of samples used forcomputing the entropy, and hence more efficient for higher dimensional problems

文献总结：

点击“阅读原文”，了解论文详情！