Local and Global Linear Convergence of General Low-Rank Matrix Recovery Problems

Authors: Yingjie Bi, Haixiang Zhang, Javad Lavaei10129-10137

AAAI 2022 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental In this section, we conduct numerical experiments to demonstrate the behavior of the perturbed gradient descent algorithm for solving low-rank matrix recovery problems.
Researcher Affiliation Academia 1 Industrial Engineering and Operations Research, University of California, Berkeley 2 Department of Mathematics, University of California, Berkeley
Pseudocode Yes Algorithm 1 in Appendix C
Open Source Code No The paper does not provide concrete access to source code for the described methodology.
Open Datasets No The paper describes data generation processes (e.g., 'each entry of Ai is independently generated from the standard Gaussian distribution', 'ground truth M = XXT or M = UV T is generated randomly') rather than using a publicly available dataset with a specific link or citation.
Dataset Splits No The paper does not provide specific dataset split information (training, validation, test) for reproducibility, as data is generated for experiments rather than split from a pre-existing dataset.
Hardware Specification No The paper does not specify any hardware details (e.g., GPU/CPU models, memory) used for running experiments.
Software Dependencies No The paper does not provide specific ancillary software details with version numbers.
Experiment Setup Yes For the symmetric problem (4), we first generate 10^4 random matrices X Rn 2r with each entry independently selected from the standard Gaussian distribution... The ground truth M = XXT or M = UV T is generated randomly with each entry of X or (U, V ) independently selected from the standard Gaussian distribution. The initial point is generated in the same way. Figure 1(a) A symmetric linear problem with r = 1, n = 40, p = 120... (b) An asymmetric linear problem with r = 5, n = 10, m = 8, p = 220... (c) The 1-bit matrix recovery problem with r = 5, n = 10. (d) The 1-bit matrix recovery problem with r = 2, n = 600.