Overview

• The package includes a demo comparing the identification results of LS, L1-LS, and NN-L1-LS.
• NN-L1-LS solves the following optimization problem

• References:
  • [1] Q. Ling, W. Shi, G. Wu, Z. Tian, “Sparse FIR Estimation of Low-Order Systems,” in Proceedings of the 2012 IEEE Statistical Signal Processing Workshop (SSP), pp. 321-324, Ann Arbor, August 2012.
  • [2] W. Shi, “FIR Estimation of Discrete MIMO Systems,” technical report, Institute of Industry Automation, University of Science and Technology of China, 2012. [Poster with numerical experiments]
• Please use the code under the GNU GPLv3
• Please note: this code is still in beta version, and is not yet well documented. The demo main file is MIMO_SI.m. The files SADMM_full.m, LADMM_full.m, DADMM_full.m, and AADMM_full.m are four different ADMM variants for solving the above NN-L1-LS problem. Numerical experiments show that the speeds: DADMM>AADMM>LADMM>SADMM. Besides, LADMM_partial solves a simple approximation of the original optimization problem thus is fast (faster than DADMM) but sometimes gives bad estimation. The code has been tested in Windows 7, PCWIN64 environment.
• Changelog:

Matlab code

• NN_L1_LS_v0.1.zip