Feature Selection vs Theory Reformulation: A Study of Genetic Refinement of Knowledge-based Neural Networks

摘要

Expert classification systems have proven themselves effective decision makers for many types of problems. However, the accuracy of such systems is often highly dependent upon the accuracy of a human expert's domain theory. When human experts learn or create a set of rules, they are subject to a number of hindrances. Most significantly experts are, to a greater or lesser extent, restricted by the tradition of scholarship which has preceded them and by an inability to examine large amounts of data in a rigorous fashion without the effects of boredom or frustration. As a result, human theories are often erroneous or incomplete. To escape this dependency, machine learning systems have been developed to automatically refine and correct an expert's domain theory. When theory revision systems are applied to expert theories, they often concentrate on the reformulation of the knowledge provided rather than on the reformulation or selection of input features. The general assumption seems to be that the expert has already selected the set of features that will be most useful for the given task. That set may, however, be suboptimal. This paper studies theory refinement and the relative benefits of applying feature selection versus more extensive theory reformulation.