Goal-Directed Classification Using Linear Machine Decision Trees

Recent work in feature-based classi cation has focused on non-parametric techniques that can classify instances even when the underlying feature distributions are unknown. The inference algorithms for training these techniques, however, are designed to maximize the accuracy of the classi er, with all errors weighted equally. In many applications, certain errors are far more costly than others, and the need arises for non-parametric classi cation techniques that can be trained to optimize task-speci c cost functions. This paper reviews the Linear Machine Decision Tree (LMDT) algorithm for inducing multivariate decision trees, and shows how LMDT can be altered to induce decision trees that minimize arbitrary misclassi cation cost functions (MCFs). Demonstrations of pixel classi cation in outdoor scenes show how MCFs can optimize the performance of embedded classi ers within the context of larger image understanding systems.