Abstract:The recently proposed Multi-Layer Convolutional Sparse Coding (ML-CSC) model, consisting of a cascade of convolutional sparse layers, provides a new interpretation of Convolutional Neural Networks (CNNs). Under this framework, the computation of the forward pass in a CNN is equivalent to a pursuit algorithm aiming to estimate the nested sparse representation vectors -- or feature maps -- from a given input signal. Despite having served as a pivotal connection between CNNs and sparse modeling, a deeper understanding of the ML-CSC is still lacking: there are no pursuit algorithms that can serve this model exactly, nor are there conditions to guarantee a non-empty model. While one can easily obtain signals that approximately satisfy the ML-CSC constraints, it remains unclear how to simply sample from the model and, more importantly, how one can train the convolutional filters from real data. In this work, we propose a sound pursuit algorithm for the ML-CSC model by adopting a projection approach. We provide new and improved bounds on the stability of the solution of such pursuit and we analyze different practical alternatives to implement this in practice. We show that the training of the filters is essential to allow for non-trivial signals in the model, and we derive an online algorithm to learn the dictionaries from real data, effectively resulting in cascaded sparse convolutional layers. Last, but not least, we demonstrate the applicability of the ML-CSC model for several applications in an unsupervised setting, providing competitive results. Our work represents a bridge between matrix factorization, sparse dictionary learning and sparse auto-encoders, and we analyze these connections in detail.

Online Convolutional Sparse Coding with Sample-Dependent Dictionary

Scalable Online Convolutional Sparse Coding.

Supervised Convolutional Sparse Coding

Laplacian Sparse Coding Dictionary for Image Set Based Collaborative Representation

Convolutional Sparse Coding for Compressed Sensing CT Reconstruction

Dictionary Learning for Image Coding Based on Multisample Sparse Representation

Generalized Convolutional Sparse Coding with Unknown Noise

K-CPD: Learning of Overcomplete Dictionaries for Tensor Sparse Coding.

Multi-Layer Convolutional Sparse Modeling: Pursuit and Dictionary Learning

Dictionary Learning for Sparse Coding: Algorithms and Convergence Analysis

Learning Multiscale Convolutional Dictionaries for Image Reconstruction

Efficient Two-Dimensional Sparse Coding Using Tensor-Linear Combination

Image Compressed Sensing Based on Multi-Level Adaptive Learning Dictionaries

Convolutional Sparse Autoencoders for Image Classification.

First and Second Order Methods for Online Convolutional Dictionary Learning

Slice-Based Online Convolutional Dictionary Learning

Dictionary Learning in Convolutional Sparse Representation

Kernel Regularized Nonlinear Dictionary Learning for Sparse Coding

Self-Explanatory Convex Sparse Representation for Image Classification

Robust Sparse Coding Via Self-Paced Learning.

Maximum Constrained Sparse Coding for Image Representation