Deep Convolutional Transform Learning -- Extended version

TL;DR

This paper presents Deep Convolutional Transform Learning (DCTL), an unsupervised deep learning method that stacks convolutional transforms to learn independent kernels, with proven convergence and superior performance over shallow methods on benchmarks.

Contribution

It introduces a novel unsupervised deep learning framework with convergence guarantees, extending shallow transform learning to multiple convolutional layers.

Findings

DCTL outperforms shallow CTL on benchmark datasets

The method converges reliably due to the proximal minimization scheme

Features learned are effective for classification and clustering

Abstract

This work introduces a new unsupervised representation learning technique called Deep Convolutional Transform Learning (DCTL). By stacking convolutional transforms, our approach is able to learn a set of independent kernels at different layers. The features extracted in an unsupervised manner can then be used to perform machine learning tasks, such as classification and clustering. The learning technique relies on a well-sounded alternating proximal minimization scheme with established convergence guarantees. Our experimental results show that the proposed DCTL technique outperforms its shallow version CTL, on several benchmark datasets.