A Supra - Classi er Architecture for Scalable

When faced with inadequate information , humans often use knowledge gained from previous experience to help them in making decisions. Even when this knowledge is spread thinly among many previous experiences, humans are able to eeectively accumulate and apply it to a current classiica-tion task of interest. Inspired by human knowledge reuse, we have previously introduced a general framework for the use of knowledge embodied in existing classi-ers to aid in a new classiication task. In this framework, a supra-classiier is built to make decisions based on the outputs of large numbers of previously trained clas-siiers designed for diierent, but possibly relevant tasks. In this article, we discuss the Hamming Nearest Neighbor (HNN) supra-classiier architecture and mathematically show its usefulness. Experiments on public domain data sets demonstrate the practicality of the framework and HNN supra-classiier when faced with very few training examples.