(c) Masaki Onuki, 2017
Author: Masaki Onuki (masaki.o@msp-lab.org)
These MATLAB codes are the examples of fast singular value shrinkage using Chebyshev polynomial approximation (CPA). When you use these samples for your paper, please cite the paper below:
M. Onuki, S. Ono, K. Shirai, and Y. Tanaka, "Fast Singular Value Shrinkage with Chebyshev Polynomial Approximation Based on Signal Sparsity," IEEE Transactions on Signal Processing.
This paper includes theoretical and practical details of the fast singular value shrinkage with CPA.
MATLAB 2015b or later. We have not confirmed yet if our codes can be run using older versions than MATLAB 2015b.
We have prepared the application indicated in Section V-F of the above paper. In the "Sample_code_Fast_SVS_CPA" folder, the singular value shrinkage using CPA and the exact method was applied to matrix rank minimization. You can select the CPA-based method or the exact method by changing the variable CPA_SVS. The CPA-based method is used if CPA_SVS = 1, or the exact method is used if CPA_SVS = 0. The matrix rank and the missing rate of the used data can be determined by changing the variables rank and Miss_rate. Additionally, choose the approximation order for CPA by changing the variable Approx order. Finally, run the source code. Enjoy!
Aug. 17, 2017: v0.1 - (original release)