An Item-Based Efficient Algorithm of Learning from Examples

Abstract

This paper presents a novel algorithm of learning from examples based on items, while the traditional algorithms could be considered as example-based. The deference is that, though both use frequency matrixes as main heuristic information , the traditional algorithm obtain the matrixes through the line-by-line scanning of examples, while the presented algorithm through scanning of items and then carrying out some calculations according to some laws found by the author in examples space. Since a item may contains a lot of examples and the cost of scanning of a example is as the same as of scanning of a item, the presented algorithm possesses great ability to be able to handle those learning tasks with great number of examples , many variables and assignments, while the traditional algorithms could not treats with those tasks.