Conditional Random Fields and Support Vector Machines: A Hybrid Approach

TL;DR

This paper introduces a hybrid loss combining CRF log loss and SVM hinge loss for structured prediction, providing theoretical conditions for Fisher consistency and demonstrating improved empirical performance.

Contribution

It presents a novel convex hybrid loss for structured prediction that unifies probabilistic and margin-based methods, with theoretical analysis and empirical validation.

Findings

Hybrid loss often outperforms individual losses in experiments.

Fisher consistency depends on label dominance measures.

Hybrid approach bridges probabilistic and margin-based prediction methods.

Abstract

We propose a novel hybrid loss for multiclass and structured prediction problems that is a convex combination of log loss for Conditional Random Fields (CRFs) and a multiclass hinge loss for Support Vector Machines (SVMs). We provide a sufficient condition for when the hybrid loss is Fisher consistent for classification. This condition depends on a measure of dominance between labels - specifically, the gap in per observation probabilities between the most likely labels. We also prove Fisher consistency is necessary for parametric consistency when learning models such as CRFs. We demonstrate empirically that the hybrid loss typically performs as least as well as - and often better than - both of its constituent losses on variety of tasks. In doing so we also provide an empirical comparison of the efficacy of probabilistic and margin based approaches to multiclass and structured prediction and the effects of label dominance on these results.