Online Decision Making by Convex Optimization

2014 Fiscal Year Final Research Report

• Project/Area Number: 23300003
• Research Category: Grant-in-Aid for Scientific Research (B)
• Allocation Type: Single-year Grants
• Section: 一般
• Research Field: Fundamental theory of informatics
• Research Institution: Kyushu University
• Principal Investigator: TAKIMOTO Eiji 九州大学, システム情報科学研究科(研究院, 教授 (50236395)
• Co-Investigator(Kenkyū-buntansha): HATANO Kohei 九州大学, システム情報科学研究院, 助教 (60404026)
• Project Period (FY): 2011-04-01 – 2015-03-31
• Keywords: 計算学習理論 / オンライン予測 / 凸最適化 / 組合せ最適化 / オンラインアルゴリズム

Outline of Final Research Achievements

Online decision making is a process where decision making and data observation are repeated. It is widely recognized that, for various problems of decision making, convex optimization theory gives a general framework of design and analysis of algorithms. However, it is limited to those problems where the domains or the decision spaces are continuous. In this study, we extend the theory so that it can be uniformly applied to a wide class of discrete domains such as rankings, spanning trees and a set of satisfying assignments. Furthermore, we consider a generalized problem where the cost of decision making for each trial depends on the past decisions made so far, and give a condition under which the problem has an efficient and high performing algorithm.

Free Research Field

計算学習理論