engleski

A relevancy filter for constructive induction

Given a set of training examples, the learner's vocabulary may be too restricted for solving the learning task, therefore numerous approaches have been proposed that allow the learner to extend its vocabulary with new terms. This work proposes a simple filter that can eliminate the constructed terms that are irrelevant for the learning task. Consequently, by biasing constructive induction (or predicate invention) to relevant terms only, the explored space of potentially relevant terms can be much larger. The elimination of irrelevant terms is specially well suited for those learning approaches with a large time or space complexity (such as genetic algorithms and neural nets).

induction; literal; relevancy

nije evidentirano