Signal and Image Processing Seminar

This work introduces the concept of classifier knowledge reuse as a means of exploiting domain knowledge from previously created, relevant classifiers to build a more accurate classifier than can be constructed using a training set and a priori knowledge alone. The supra-classifier framework is a practical approach to reuse that is both flexible in the type of classifier reused and scalable in the number of classifiers reused. Supra-classifier design issues are explored and several supra-classifier architectures are described and compared in the paradigm of a knowledge space. The Hamming Nearest Neighbor supra-classifier is given particular treatment and is both theoretically and empirically shown to be a useful supra-classifier architecture. An application of knowledge reuse to a classifier knowledge base is introduced and the effectiveness of the supra-classifier framework to this real world application is empirically demonstrated.