The original paper is in English. Non-English content has been machine-translated and may contain typographical errors or mistranslations. ex. Some numerals are expressed as "XNUMX".
Copyrights notice
The original paper is in English. Non-English content has been machine-translated and may contain typographical errors or mistranslations. Copyrights notice
A filtragem adaptativa multikernel é uma abordagem não linear atraente para tarefas de estimativa/rastreamento online. Apesar de suas vantagens potenciais sobre sua contraparte de kernel único, o uso de kernels ponderados inadequadamente pode resultar em um ganho de desempenho insignificante. Neste artigo, propomos uma técnica eficiente de ponderação recursiva de kernel para filtragem adaptativa multikernel para ativar todos os kernels. Os pesos propostos equalizam as taxas de convergência de todos os erros dos coeficientes parciais correspondentes. Os pesos propostos são implementados através de um determinado desenho métrico baseado na matriz de ponderação. Exemplos numéricos mostram, para conjuntos de dados reais sintéticos e múltiplos, que a técnica proposta apresenta um desempenho melhor do que os pesos de kernel ajustados manualmente e que supera significativamente o algoritmo de regressão de kernel múltiplo online.
Kwangjin JEONG
Keio University
Masahiro YUKAWA
Keio University
The copyright of the original papers published on this site belongs to IEICE. Unauthorized use of the original or translated papers is prohibited. See IEICE Provisions on Copyright for details.
Copiar
Kwangjin JEONG, Masahiro YUKAWA, "Kernel Weights for Equalizing Kernel-Wise Convergence Rates of Multikernel Adaptive Filtering" in IEICE TRANSACTIONS on Fundamentals,
vol. E104-A, no. 6, pp. 927-939, June 2021, doi: 10.1587/transfun.2020EAP1080.
Abstract: Multikernel adaptive filtering is an attractive nonlinear approach to online estimation/tracking tasks. Despite its potential advantages over its single-kernel counterpart, a use of inappropriately weighted kernels may result in a negligible performance gain. In this paper, we propose an efficient recursive kernel weighting technique for multikernel adaptive filtering to activate all the kernels. The proposed weights equalize the convergence rates of all the corresponding partial coefficient errors. The proposed weights are implemented via a certain metric design based on the weighting matrix. Numerical examples show, for synthetic and multiple real datasets, that the proposed technique exhibits a better performance than the manually-tuned kernel weights, and that it significantly outperforms the online multiple kernel regression algorithm.
URL: https://global.ieice.org/en_transactions/fundamentals/10.1587/transfun.2020EAP1080/_p
Copiar
@ARTICLE{e104-a_6_927,
author={Kwangjin JEONG, Masahiro YUKAWA, },
journal={IEICE TRANSACTIONS on Fundamentals},
title={Kernel Weights for Equalizing Kernel-Wise Convergence Rates of Multikernel Adaptive Filtering},
year={2021},
volume={E104-A},
number={6},
pages={927-939},
abstract={Multikernel adaptive filtering is an attractive nonlinear approach to online estimation/tracking tasks. Despite its potential advantages over its single-kernel counterpart, a use of inappropriately weighted kernels may result in a negligible performance gain. In this paper, we propose an efficient recursive kernel weighting technique for multikernel adaptive filtering to activate all the kernels. The proposed weights equalize the convergence rates of all the corresponding partial coefficient errors. The proposed weights are implemented via a certain metric design based on the weighting matrix. Numerical examples show, for synthetic and multiple real datasets, that the proposed technique exhibits a better performance than the manually-tuned kernel weights, and that it significantly outperforms the online multiple kernel regression algorithm.},
keywords={},
doi={10.1587/transfun.2020EAP1080},
ISSN={1745-1337},
month={June},}
Copiar
TY - JOUR
TI - Kernel Weights for Equalizing Kernel-Wise Convergence Rates of Multikernel Adaptive Filtering
T2 - IEICE TRANSACTIONS on Fundamentals
SP - 927
EP - 939
AU - Kwangjin JEONG
AU - Masahiro YUKAWA
PY - 2021
DO - 10.1587/transfun.2020EAP1080
JO - IEICE TRANSACTIONS on Fundamentals
SN - 1745-1337
VL - E104-A
IS - 6
JA - IEICE TRANSACTIONS on Fundamentals
Y1 - June 2021
AB - Multikernel adaptive filtering is an attractive nonlinear approach to online estimation/tracking tasks. Despite its potential advantages over its single-kernel counterpart, a use of inappropriately weighted kernels may result in a negligible performance gain. In this paper, we propose an efficient recursive kernel weighting technique for multikernel adaptive filtering to activate all the kernels. The proposed weights equalize the convergence rates of all the corresponding partial coefficient errors. The proposed weights are implemented via a certain metric design based on the weighting matrix. Numerical examples show, for synthetic and multiple real datasets, that the proposed technique exhibits a better performance than the manually-tuned kernel weights, and that it significantly outperforms the online multiple kernel regression algorithm.
ER -