Seminar on Computational Learning and Adaptation


  Learning Taxonomic Relations by Case-based Reasoning

Ken Satoh
Professor, Foundations of Information Research Division
National Institute of Informatics, Japan
ksatoh@nii.ac.jp

In this paper, we propose a learning method that requires a minimal case base to represent taxonomic relations in a tree-structured concept hierarchy. We firstly propose case-based taxonomic reasoning and show an upper bound on the necessary positive cases and negative cases to represent a relation. Then, we describe a learning method that requires a minimal case base, with sampling and membership queries. We analyze this learning method by sample complexity and query complexity in the framework of PAC learning. This is an extension of work on approximating a critical case base which represents a boolean function using set-inclusion based similarity. The work is related to "monotone theory", developed by Bshouty, Kardon and Roth.



Date: Thursday, February 21

Time: 4:15-5:30PM

Place: Cordura 100


Return to the seminar schedule