Gradient descent with sparsification

2009166 citationsJournal Article

Authors

Rahul Garg · IBM Research - Thomas J. Watson Research Center

Rohit Khandekar · IBM Research - Thomas J. Watson Research Center

Abstract

We present an algorithm for finding an s-sparse vector x that minimizes the square-error ∥y -- Φx∥2 where Φ satisfies the restricted isometry property (RIP), with isometric constant δ2s < 1/3. Our algorithm, called GraDeS (Gradient Descent with Sparsification) iteratively updates x as: [EQUATION]

Topics & Keywords

Sparse and Compressive Sensing Techniques Stochastic Gradient Optimization Techniques Advanced Optimization Algorithms Research

UN Sustainable Development Goals

Reduced inequalities

Publication Details

DOI: 10.1145/1553374.1553417

Field-Weighted Citation Impact: 16.46