Max-margin Deep Generative Models

TL;DR

This paper introduces max-margin deep generative models (mmDGMs) that enhance the discriminative capabilities of DGMs using max-margin principles while maintaining their generative abilities, demonstrated on MNIST and SVHN datasets.

Contribution

The paper proposes a novel max-margin framework for DGMs, combining discriminative max-margin learning with generative modeling, and develops an efficient optimization algorithm.

Findings

Max-margin DGMs improve prediction accuracy over traditional DGMs.

mmDGMs retain strong generative capabilities.

Competitive performance with state-of-the-art discriminative networks.

Abstract

Deep generative models (DGMs) are effective on learning multilayered representations of complex data and performing inference of input data by exploring the generative ability. However, little work has been done on examining or empowering the discriminative ability of DGMs on making accurate predictions. This paper presents max-margin deep generative models (mmDGMs), which explore the strongly discriminative principle of max-margin learning to improve the discriminative power of DGMs, while retaining the generative capability. We develop an efficient doubly stochastic subgradient algorithm for the piecewise linear objective. Empirical results on MNIST and SVHN datasets demonstrate that (1) max-margin learning can significantly improve the prediction performance of DGMs and meanwhile retain the generative ability; and (2) mmDGMs are competitive to the state-of-the-art fully discriminative networks by employing deep convolutional neural networks (CNNs) as both recognition and generative models.