Maximum Entropy Discrimination Markov Networks

TL;DR

This paper introduces MaxEnDNet, a unified framework combining Bayesian estimation and max-margin structured learning for Markov networks, leading to improved inference and generalization in structured prediction tasks.

Contribution

It presents a novel Bayesian-structured max-margin learning framework for Markov networks, extending existing models and providing efficient inference and generalization guarantees.

Findings

Outperforms existing structured learning methods on synthetic data

Provides a sparse, regularized Markov network model

Offers a PAC-Bayesian generalization bound

Abstract

In this paper, we present a novel and general framework called {\it Maximum Entropy Discrimination Markov Networks} (MaxEnDNet), which integrates the max-margin structured learning and Bayesian-style estimation and combines and extends their merits. Major innovations of this model include: 1) It generalizes the extant Markov network prediction rule based on a point estimator of weights to a Bayesian-style estimator that integrates over a learned distribution of the weights. 2) It extends the conventional max-entropy discrimination learning of classification rule to a new structural max-entropy discrimination paradigm of learning the distribution of Markov networks. 3) It subsumes the well-known and powerful Maximum Margin Markov network (M$^3$N) as a special case, and leads to a model similar to an $L_1$-regularized M$^3$N that is simultaneously primal and dual sparse, or other types of Markov network by plugging in different prior distributions of the weights. 4) It offers a simple inference algorithm that combines existing variational inference and convex-optimization based M$^3$N solvers as subroutines. 5) It offers a PAC-Bayesian style generalization bound. This work represents the first successful attempt to combine Bayesian-style learning (based on generative models) with structured maximum margin learning (based on a discriminative model), and outperforms a wide array of competing methods for structured input/output learning on both synthetic and real data sets.