Information-theoretic Dictionary Learning for Image Classification

TL;DR

This paper introduces an information-theoretic method for learning compact, discriminative, and generative dictionaries to improve image classification accuracy, using a two-stage mutual information maximization approach.

Contribution

It proposes a novel two-stage dictionary learning algorithm based on mutual information maximization, enhancing discriminative and reconstructive capabilities for image classification.

Findings

Effective in selecting compact and discriminative dictionaries

Improves classification accuracy on real datasets

Demonstrates superiority over existing methods

Abstract

We present a two-stage approach for learning dictionaries for object classification tasks based on the principle of information maximization. The proposed method seeks a dictionary that is compact, discriminative, and generative. In the first stage, dictionary atoms are selected from an initial dictionary by maximizing the mutual information measure on dictionary compactness, discrimination and reconstruction. In the second stage, the selected dictionary atoms are updated for improved reconstructive and discriminative power using a simple gradient ascent algorithm on mutual information. Experiments using real datasets demonstrate the effectiveness of our approach for image classification tasks.