Geometric Models with Co-occurrence Groups

TL;DR

This paper introduces a geometric model for sparse signal representations optimized via co-occurrence groups and Bernoulli mixture models, demonstrating effectiveness in face image compression and digit classification.

Contribution

It presents a novel geometric modeling approach that leverages co-occurrence groups and maximum likelihood estimation for sparse signals.

Findings

Effective face image compression demonstrated

Accurate MNIST digit classification achieved

Model outperforms traditional sparse representation methods

Abstract

A geometric model of sparse signal representations is introduced for classes of signals. It is computed by optimizing co-occurrence groups with a maximum likelihood estimate calculated with a Bernoulli mixture model. Applications to face image compression and MNIST digit classification illustrate the applicability of this model.