Designing large scale classifiers
William A. Porter, Wei Liu · 2002
In this study we present a design for hierarchical modular classifiers. The design features an algorithm which selects a set of exemplars. Using these exemplars the classification problem is decomposed into a family of disjoint subproblems. A classification module is trained for each subproblem. The collection of classification modules and a rule book for their use then comprise the resultant design.