This repository is our implementation of
In this paper, we introduce a new surrogate for matrix completion, which is equivalent to the nuclear norm.
In particular, we prove the upper-bound of an approximate/inexact closed-form solution, which is a crucial step of the optimization. The surrogate and its optimization make the matrix completion more compatible for additional learning mechanisms.
If you have issues, please email:
hyzhang98@gmail.com or hyzhang98@mail.nwpu.edu.cn.
- image_main.m: An example of how to run the code
- ncarl_no_noise: The source code of NCARL
- ncarl.m: The noisy extension
@article{NCARL,
author={Li, Xuelong and Zhang, Hongyuan and Zhang, Rui},
journal={IEEE Transactions on Pattern Analysis and Machine Intelligence},
title={Matrix Completion via Non-Convex Relaxation and Adaptive Correlation Learning},
year={2022},
volume={},
number={},
pages={1-1},
doi={10.1109/TPAMI.2022.3157083}
}