Local discriminative learning for pattern recognition

Local discriminative learning methods approximate a target function (a post eriori class probability function) directly by partitioning the feature spa ce into a set of local regions, and appropriately modeling a simple input-o utput relationship (function) in each one. This paper presents a new method for judiciously partitioning the input feature space in order to accuratel y represent the target function. The method accomplishes this by approximat ing not only the target function itself but also its derivatives. As such, the method partitions the input feature space along those dimensions for wh ich the class probability function changes most rapidly, thus minimizing bi as. The efficacy of the method is validated using a variety of simulated an d real-world data. (C) 2000 Pattern Recognition Society. Published by Elsev ier Science Ltd. All rights reserved.