Matrix recovery with the awareness of covariate information and variable selection
by Yang Qing
THESIS
2019
M.Phil. Mathematics
xi, 44 pages ; 30 cm
Abstract
In this thesis, we propose the variable selection process in the matrix recovery task with the consideration of covariate information, by shrinking the size
of factors used in predicting unobserved matrix entries. Matrix completion has
attracted a wide range of attention in recent years, where there are various applications including: recommender system, image processing and video denoising.
Here we review the existing matrix completion literature, and further discuss
the current methods who have considered the improvement contributed by additional covariate information. Expanding current matrix recovery method with
covariate information, we implement variable selection to factors assisted and
further improve the prediction performance.
Post a Comment