The Projected Belief Network Classfier : both Generative and Discriminative

TL;DR

This paper introduces a convolutional projected belief network that is both fully generative and discriminative, demonstrating strong performance and data synthesis capabilities on spoken command spectrograms.

Contribution

It presents a novel convolutional PBN that combines generative and discriminative features within a single model, tested on speech spectrograms.

Findings

Excellent generative and discriminative qualities observed.

Effective data synthesis and reconstruction demonstrated.

Classifier performance close to traditional discriminative networks.

Abstract

The projected belief network (PBN) is a layered generative network with tractable likelihood function, and is based on a feed-forward neural network (FF-NN). It can therefore share an embodiment with a discriminative classifier and can inherit the best qualities of both types of network. In this paper, a convolutional PBN is constructed that is both fully discriminative and fully generative and is tested on spectrograms of spoken commands. It is shown that the network displays excellent qualities from either the discriminative or generative viewpoint. Random data synthesis and visible data reconstruction from low-dimensional hidden variables are shown, while classifier performance approaches that of a regularized discriminative network. Combination with a conventional discriminative CNN is also demonstrated.