LMS Algorithms for Multilinear Forms

• Published in: 2020 International Symposium on Electronics and Telecommunications (ISETC) pp. 1 - 4
• Main Authors: Dogariu, Laura-Maria, Paleologu, Constantin, Benesty, Jacob, Oprea, Cristina, Ciochina, Silviu
• Format: Conference Proceeding
• Language: English
• Published: IEEE 05-11-2020
• Subjects: Adaptive filters; Complexity theory; Convergence; Estimation; Indexes; LMS algorithm; MISO communication; multilinear forms; NLMS algorithm; Optimization; Simulation; system identification
• ISBN: 1728189217; 9781728189215
• ISSN: 2475-7861
• DOI: 10.1109/ISETC50328.2020.9301133

Solving a high-dimension system identification problem could involve significant challenges in terms of complexity and accuracy of the solution. Due to the large parameter space, a decomposition-based approach fits very well in this context. This was the idea behind the recently developed iterative Wiener filter for multilinear forms, which reformulates the problem using a combination of shorter filters. Nevertheless, there are inherent limitations related to the Wiener solution, while the least-mean-square (LMS) adaptive filter would represent a more practical alternative. Consequently, in this paper, we develop LMS-based algorithms for multilinear forms, in the context of a multiple-input/single-output system identification problem. Simulation results indicate the good performance of the proposed algorithms, especially in terms of their fast convergence features.