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