単一・マルチタスク最適化のための最小後悔探索
Minimum Regret Search for Single- and Multi-Task Optimization
ベイズ最適化の新たな獲得関数「最小後悔探索(MRS)」を提案し、最終推薦の期待単純後悔を最小化することで、エントロピー探索より外れ値が少ないことを合成問題とマルチタスクロボット制御で示した。
分類: stat.ML, cs.IT, cs.LG, cs.RO, math.IT
原文アブストラクト
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.
