Tarm: A Turbo-Type Algorithm for Low-Rank Matrix Recovery

Zhipeng Xue; Xiaojun Yuan; Junjie Ma

ICASSP 2018accepted0 citations

Tarm: A Turbo-Type Algorithm for Low-Rank Matrix Recovery

Abstract

This paper is concerned with the affine rank minimization (ARM) problem for low-rank matrix recovery purposes. Inspired by the recently proposed Turbo-CS algorithm in the field of compressed sensing, we propose a turbo-type algorithm for ARM, termed Turbo-ARM (TARM). For matrix recovery problems with a large class of random measurement matrices, the performance of TARM can be analyzed via the state evolution framework. Our numerical results show that TARM achieves state-of-the-art reconstruction performance, and our results are further confirmed by state evolution analysis.

BibTeX

@inproceedings{icassp2018_tarmaturbotypeal,
  title = {Tarm: A Turbo-Type Algorithm for Low-Rank Matrix Recovery},
  author = {Zhipeng Xue and Xiaojun Yuan and Junjie Ma},
  booktitle = {ICASSP 2018},
  year = {2018}
}