Sparsity by Redundancy: Solving $L_1$ with a Simple Reparametrization

Liu Ziyin, Zihao Wang

Keywords: L1, sparsity, feature selection, deep learning, landscape

Abstract: We identify and prove a general principle: $L_1$ sparsity can be achieved using a redundant parametrization plus $L_2$ penalty. Our results lead to a simple algorithm, \textit{spred}, that seamlessly integrates $L_1$ regularization into any modern deep learning framework. Practically, we demonstrate (1) the efficiency of \textit{spred} in optimizing conventional tasks such as lasso and sparse coding, (2) benchmark our method for nonlinear feature selection of six gene selection tasks, and (3) illustrate the usage of the method for achieving structured and unstructured sparsity in deep learning in an end-to-end manner. Conceptually, our result bridges the gap in understanding the inductive bias of the redundant parametrization common in deep learning and conventional statistical learning.

Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors’ identity.

No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.

Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics

Submission Guidelines: Yes

Please Choose The Closest Area That Your Submission Falls Into: Deep Learning and representational learning

TL;DR: We propose a simple method for solving general L1 penalized objectives with gradient descent

Community Implementations: [ 2 code implementations](https://www.catalyzex.com/paper/arxiv:2210.01212/code)