情報論的学習理論(<論文特集>「情報論的学習理論(IBIS2000)」)

抄録

This paper overviews researchs in Information-Based Induction Sciences(IBIS).IBIS is a very inter-disciplinary research movement based on theoretical frameworks and techniques originated from various areas such as information theory, statistical learning theory, stayistics, statistical mechanics, and information geometry.Applications of theoretical results and methods to real-world problems are also within the scope.In order to consistently review researches in such wide area, we first introduce a simple but general formulation of learning problem established in statistical learning theory, and explain various approaches along the formulation.