PubAg

Main content area

Empirical Bayes matrix completion

Author:
Matsuda, Takeru, Komaki, Fumiyasu
Source:
Computational statistics & data analysis 2019 v.137 pp. 195-210
ISSN:
0167-9473
Subject:
algorithms, models, shrinkage, variance
Abstract:
We develop an empirical Bayes (EB) algorithm for the matrix completion problems. The EB algorithm is motivated from the singular value shrinkage estimator for matrix means by Efron and Morris. Since the EB algorithm is derived as the Expectation–Maximization algorithm applied to a simple model, it does not require heuristic parameter tuning other than tolerance. Also, it can account for the heterogeneity in variance of observation noise. Numerical results demonstrate that the EB algorithm attains at least comparable accuracy to existing algorithms for matrices not close to square and that it works particularly well when the rank is relatively large or the proportion of observed entries is small. Application to real data also shows the practical utility of the EB algorithm.
Agid:
6333510