Skip Navigation
York U: Redefine the PossibleHOME | Current Students | Faculty & Staff | Research | International
Search »FacultiesLibrariesCampus MapsYork U OrganizationDirectorySite Index
Future Students, Alumni & Visitors
2004 Technical Reports

Discriminative Training for Large Margin HMMs

Hui Jiang

Technical Report CS-2004-01

York University

March 30, 2004

Abstract

In this report, motivated by large margin classifiers in machine learning, we propose a new discriminative training criterion for estimating CDHMM (continuous density hidden Markov model) in speech recognition based on the principle of maximizing the minimum multi-class separation margin. In this report, we first show that this maximum margin model estimation problem can be formulated as a standard constrained minimax optimization problem. Alternatively, we also show that the estimation problem can be solved by a GPD (generalized probabilistic descent) algorithm if we approximate the objective function by a continuous and differentiable function, such as summation of exponential functions. In this report, we also propose a method to handle classification errors in training set in maximum margin estimation by using them to optimize a separate objective function which is similar to that in the MCE (minimum classification error) formulation.

Download paper in PDF format.



The documents distributed by this server have been provided by the contributing authors as a means to ensure timely dissemination of scholarly and technical work on a noncommercial basis. Copyright and all rights therein are maintained by the authors or by other copyright holders, notwithstanding that they have offered their works here electronically. It is understood that all persons copying this information will adhere to the terms and constraints invoked by each author's copyright. These works may not be reposted without the explicit permission of the copyright holder.