Convolutional Deep Exponential Families

TL;DR

This paper introduces convolutional deep exponential families (CDEFs), a hierarchical probabilistic model that reduces parameters and effectively captures temporal correlations with limited data.

Contribution

The paper proposes CDEFs, combining deep exponential families with convolutional structures to improve parameter efficiency and temporal correlation modeling.

Findings

CDEFs effectively uncover time correlations.

CDEFs require less data to learn hierarchical dependencies.

Parameter tying reduces model complexity.

Abstract

We describe convolutional deep exponential families (CDEFs) in this paper. CDEFs are built based on deep exponential families, deep probabilistic models that capture the hierarchical dependence between latent variables. CDEFs greatly reduce the number of free parameters by tying the weights of DEFs. Our experiments show that CDEFs are able to uncover time correlations with a small amount of data.