Multiplication-Avoiding Variant of Power Iteration with Applications

Abstract

Power iteration is a fundamental algorithm in data analysis. It extracts the eigenvector corresponding to the largest eigenvalue of a given matrix. Applications include ranking algorithms, recommendation systems, principal component analysis (PCA), among many others. In this paper, we introduce multiplication-avoiding power iteration (MAPI), which replaces the standard _2-inner products that appear at the regular power iteration (RPI) with multiplication-free vector products which are Mercer-type kernel operations related with the _1 norm. Precisely, for an n n matrix, MAPI requires n multiplications, while RPI needs n^2 multiplications per iteration. Therefore, MAPI provides a significant reduction of the number of multiplication operations, which are known to be costly in terms of energy consumption. We provide applications of MAPI to PCA-based image reconstruction as well as to graph-based ranking algorithms. When compared to RPI, MAPI not only typically converges much faster, but also provides superior performance.

Tasks

Reproductions