Generative-Discriminative Variational Model for Visual Recognition

TL;DR

This paper introduces a Generative-Discriminative Variational Model (GDVM) that combines generative and discriminative approaches to improve visual classification tasks like multi-class, multi-label, and zero-shot learning.

Contribution

The paper proposes a novel GDVM that integrates generative modeling with discriminative classification to enhance visual recognition performance.

Findings

GDVM outperforms baseline models in various classification tasks.

The model demonstrates improved generalization in zero-shot learning.

Generative capabilities aid in reducing overfitting.

Abstract

The paradigm shift from shallow classifiers with hand-crafted features to end-to-end trainable deep learning models has shown significant improvements on supervised learning tasks. Despite the promising power of deep neural networks (DNN), how to alleviate overfitting during training has been a research topic of interest. In this paper, we present a Generative-Discriminative Variational Model (GDVM) for visual classification, in which we introduce a latent variable inferred from inputs for exhibiting generative abilities towards prediction. In other words, our GDVM casts the supervised learning task as a generative learning process, with data discrimination to be jointly exploited for improved classification. In our experiments, we consider the tasks of multi-class classification, multi-label classification, and zero-shot learning. We show that our GDVM performs favorably against the baselines or recent generative DNN models.