Deep Dictionary Learning: A PARametric NETwork Approach

TL;DR

This paper introduces a hierarchical deep dictionary learning method for image classification that improves accuracy, robustness to adversarial attacks, and ease of tuning by learning multiple dictionaries at different scales.

Contribution

It proposes a novel hierarchical dictionary learning framework trained with classification objectives, enhancing feature extraction and robustness over existing methods.

Findings

Improved classification accuracy on four benchmark datasets.

Lower fooling rate against adversarial perturbations.

Easier tuning and adaptation with added layers.

Abstract

Deep dictionary learning seeks multiple dictionaries at different image scales to capture complementary coherent characteristics. We propose a method for learning a hierarchy of synthesis dictionaries with an image classification goal. The dictionaries and classification parameters are trained by a classification objective, and the sparse features are extracted by reducing a reconstruction loss in each layer. The reconstruction objectives in some sense regularize the classification problem and inject source signal information in the extracted features. The performance of the proposed hierarchical method increases by adding more layers, which consequently makes this model easier to tune and adapt. The proposed algorithm furthermore, shows remarkably lower fooling rate in presence of adversarial perturbation. The validation of the proposed approach is based on its classification performance using four benchmark datasets and is compared to a CNN of similar size.