Max-Margin Markov Networks 论文

摘要

3) networks incorporate both kernels, which efficiently deal with highdimensional features, and the ability to capture correlations in structured data.We present an efficient algorithm for learning M 3 networks based on a compact quadratic program formulation. We provide a new theoretical bound for general-ization in structured domains. Experiments on the task of handwritten character recognition, demonstrate very significant gains over previous approaches. 1 Introduction In supervised classification, our goal is to classify instances into some set of discrete cat-egories. Recently, support vector machines (SVMs) have demonstrated impressive successes on a broad range of tasks, including document classification, character recognition,image recognition, and many more. SVMs owe a great part of their success to their ability to use kernels, allowing the classifier to exploit a very high-dimensional (possibly eveninfinite-dimensional) feature space. In addition to their empirical success, SVMs are also appealing due to the existence of strong generalization guarantees, derived from the margin-maximizing properties of the learning algorithm.