Fixed Point Iteration Algorithms for Low-rank Matrix Completion

Statistics

Huang, Xingliang

University of Waterloo

关键词: Statistics;   

Others  :  https://uwspace.uwaterloo.ca/bitstream/10012/9370/3/Huang_Xingliang.pdf

瑞士|英语

来源: UWSPACE Waterloo Institutional Repository

【 摘 要 】

A lot of applications can be formulated as matrix completion problems. In order toaddress such problems, a common assumption is that the underlying matrix is (approximately)low-rank. Under certain conditions, the recovery of low-rank matrix can be donevia nuclear norm minimization, a convex program.Scalable and fast algorithms are essential as the practical matrix completion tasks alwaysoccur on a large scale. Here we study two algorithms and generalize the uni edframework ofxed point iteration algorithm. We derive the convergence results and proposea new algorithm based on the insights. Compared with the baseline algorithms, ourproposed method is signi cantly more e cient without loss of precision and accelerationpotentiality.iii