Deep Transform and Metric Learning Networks

TL;DR

This paper introduces a novel deep dictionary learning method that combines linear layers with recurrent neural networks to learn deep transforms and metrics, outperforming existing methods and CNNs.

Contribution

It proposes a new deep dictionary learning framework where each layer is a combination of linear and recurrent neural network components, providing new insights and improved performance.

Findings

Outperforms existing Deep DL methods

Surpasses state-of-the-art CNNs in experiments

Unveils new connections between neural networks and deep dictionary learning

Abstract

Based on its great successes in inference and denosing tasks, Dictionary Learning (DL) and its related sparse optimization formulations have garnered a lot of research interest. While most solutions have focused on single layer dictionaries, the recently improved Deep DL methods have also fallen short on a number of issues. We hence propose a novel Deep DL approach where each DL layer can be formulated and solved as a combination of one linear layer and a Recurrent Neural Network, where the RNN is flexibly regraded as a layer-associated learned metric. Our proposed work unveils new insights between the Neural Networks and Deep DL, and provides a novel, efficient and competitive approach to jointly learn the deep transforms and metrics. Extensive experiments are carried out to demonstrate that the proposed method can not only outperform existing Deep DL, but also state-of-the-art generic Convolutional Neural Networks.