Accession Number : ADA187723

Title :   Two New Frameworks for Learning.

Descriptive Note : Technical rept.,

Corporate Author : CARNEGIE-MELLON UNIV PITTSBURGH PA ROBOTICS INST

Personal Author(s) : Natarajan, B K

PDF Url : ADA187723

Report Date : Nov 1987

Pagination or Media Count : 17

Abstract : This paper presents two new formal frameworks for learning. The first framework requires the learner to approximate an unknown function, given examples for the function as well as some background information on it. It is shown that this framework is no more powerful than a framework that allows the learner to see examples but not background information. The second framework explores learning in the sense of improving computational efficiency as opposed to acquiring an unknown concept or function. Specifically, the framework concerns the acquisition of heuristics for examples over problem domains of special structure. A theorem is proved identifying some conditions sufficient to allow the efficient acquisition of heuristics over the aforementioned class of domains.

Descriptors :   *COMPUTATIONS, *LEARNING, *ALGORITHMS, ACQUISITION, DOMAIN WALLS, EFFICIENCY, HEURISTIC METHODS, LEARNING MACHINES, FUNCTIONS(MATHEMATICS), ARTIFICIAL INTELLIGENCE, ARCHITECTURE, INFORMATION THEORY

Subject Categories : Theoretical Mathematics

Distribution Statement : APPROVED FOR PUBLIC RELEASE