Robust variable-regularized RLS algorithms

Elisei-Iliescu, Camelia; Stanciu, Cristian; Paleologu, Constantin; Benesty, Jacob; Anghel, Cristian; Ciochină, Silviu

doi:10.1109/hscma.2017.7895584

Cited by 14 publications

(12 citation statements)

References 13 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…A larger value of δ will tend to slow down (or even freeze) the filter updates, which is preferable in low SNR conditions. [11][12][13] The update equation of the regularized RLS can be rewritten as:…”

Section: Regularized Rls Algorithmmentioning

confidence: 99%

See 1 more Smart Citation

Low-complexity data-reuse RLS algorithm with increased robustness features

Fîciu

Stanciu²,

Elisei-Iliescu³

et al. 2023

Advanced Topics in Optoelectronics, Microelectronics, and Nanotechnologies XI

Self Cite

View full text Add to dashboard Cite

Recent advancements in the field of adaptive filters based on the least-squares minimization criterion propose a stable and efficient recursive least-squares (RLS) algorithm with attractive numerical properties and performance similar to the classical RLS methods. The combination between the RLS and the dichotomous coordinate descent (DCD) iterations (i.e., the RLS-DCD) offers an attractive option for practical applications with low requirements in terms of chip area. This paper employs a data-reuse methodology to further improve the tracking speed of the RLS-DCD algorithm. Furthermore, a regularization principle is used to develop its robustness features in low signal-to-noise working conditions.

show abstract

Section: Regularized Rls Algorithmmentioning

confidence: 99%

“…We propose to combine the regularized RLS-DCD algorithm 12 with a DR method that can generate improved tracking capabilities. 5 We will call the new adaptive algorithm the R-DR-RLS-DCD.…”

Section: Regularized Data-reuse Rls-dcdmentioning

confidence: 99%

Low-complexity data-reuse RLS algorithm with increased robustness features

Fîciu

Stanciu²,

Elisei-Iliescu³

et al. 2023

Advanced Topics in Optoelectronics, Microelectronics, and Nanotechnologies XI

Self Cite

View full text Add to dashboard Cite

show abstract

“…The importance of having a variable regularization has been earlier recognized in other scenarios where the excitation power drops significantly [15], and in adaptive beamforming applications [16], where the sample covariance is naturally rank defficient. Recently, the DCD mechanism has been deployed for setting a varying-strength regularizer in the context of echo cancellation [17]. In this same respect, the reweighted LS problem is another example where time-varying regularization has been tackled instead via conjugate gradient techniques [2], addressed, e.g., in [18], [19].…”

Section: A Related Workmentioning

confidence: 99%

Optimal Diffusion Learning Over Networks—Part II: Multitask Algorithms

Merched

2022

IEEE Open J. Signal Process.

View full text Add to dashboard Cite

“…Horita et al (2004) use a rank-one update technique of eigendecomposition, and Gay (1996) regularizes random one dimension at each step. Elisei-Iliescu et al (2017) uses a dichotomous coordinate descent method. These techniques can be performed in O(n 2 ) time per step, but their effectiveness is sensitive to the selection of coordinates.…”

Section: Related Workmentioning

confidence: 99%

Second Order Techniques for Learning Time-series with Structural Breaks

Osogami

2021

AAAI

View full text Add to dashboard Cite

We study fundamental problems in learning nonstationary time-series: how to effectively regularize time-series models and how to adaptively tune forgetting rates. The effectiveness of L2 regularization depends on the choice of coordinates, and the variables need to be appropriately normalized. In nonstationary environment, however, what is appropriate can vary over time. Proposed regularization is invariant to the invertible linear transformation of coordinates, eliminating the necessity of normalization. We also propose an ensemble learning approach to adaptively tuning the forgetting rate and regularization-coefficient. We train multiple models with varying hyperparameters and evaluate their performance by the use of multiple hyper forgetting rates. At each step, we choose the best performing model on the basis of the best performing hyper forgetting rate. The effectiveness of the proposed approaches is demonstrated with real time-series.

show abstract

Robust variable-regularized RLS algorithms

Cited by 14 publications

References 13 publications

Low-complexity data-reuse RLS algorithm with increased robustness features

Low-complexity data-reuse RLS algorithm with increased robustness features

Optimal Diffusion Learning Over Networks—Part II: Multitask Algorithms

Second Order Techniques for Learning Time-series with Structural Breaks

Contact Info

Product

Resources

About