21
2

Approximate Guarantees for Dictionary Learning

Abstract

In the dictionary learning (or sparse coding) problem, we are given a collection of signals (vectors in Rd\mathbb{R}^d), and the goal is to find a "basis" in which the signals have a sparse (approximate) representation. The problem has received a lot of attention in signal processing, learning, and theoretical computer science. The problem is formalized as factorizing a matrix X(d×n)X (d \times n) (whose columns are the signals) as X=AYX = AY, where AA has a prescribed number mm of columns (typically mnm \ll n), and YY has columns that are kk-sparse (typically kdk \ll d). Most of the known theoretical results involve assuming that the columns of the unknown AA have certain incoherence properties, and that the coefficient matrix YY has random (or partly random) structure. The goal of our work is to understand what can be said in the absence of such assumptions. Can we still find AA and YY such that XAYX \approx AY? We show that this is possible, if we allow violating the bounds on mm and kk by appropriate factors that depend on kk and the desired approximation. Our results rely on an algorithm for what we call the threshold correlation problem, which turns out to be related to hypercontractive norms of matrices. We also show that our algorithmic ideas apply to a setting in which some of the columns of XX are outliers, thus giving similar guarantees even in this challenging setting.

View on arXiv
Comments on this paper