《03 - Incremental Learning from Noisy Data》.pdf

Machine Learning 1: 317-354, 1986 © 1986 Kluwer Academic Publishers, Boston - Manufactured in The Netherlands Incremental Learning from Noisy Data JEFFREY C. SCHLIMMER RICHARD H . GRANGER, JR . (SCHLIMMER@ICS.UCI. EDU) (GRANGER@ICS.UCI .EDU) Irvine Computational Intelligence Project, Department of Information and Computer Science, University of California, Irvine, CA 92717, U.S.A. (Received March 5, 1986) (Revised May 2, 1986) Key words: learning from examples, contingency, systematic noise, concept drift, constructive induction Abstract. Induction of a concept description given noisy instances is difficult and is further exacerbated when the concepts may change over time. This paper presents a solution which has been guided by psychological and mathematical results. The method is based on a distributed concept description which is composed of a set of weighted, symbolic characterizations. Two learning processes incrementally modify this description. One adjusts the characterization weights and another creates new characteriza- tions. The latter process is described in terms of a search through the space of possibilities and is shown to require linear space with respect to the number of attribute-value pairs in the description language. The method utilizes previously acquired concept definitions in subsequent learning by adding an attribute for each learned concept to instance descriptions. A program called STAGGER fully embodies this method, and this paper reports on a number of empirical analyses of its performance . Since un