Решение проблемы формирования устойчивых и состоятельных оценок корреляционной матрицы наблюдений методом динамической регуляризации / Скачков В. В., Чепкий В. В., Ефимчиков А. Н., Коркин А. Ю., Гончарук А. А. (2021)
Ukrainian

English  Cybernetics and Systems Analysis   /     Issue (2021, 57 (1))

Skachkov V.V., Chepkii V.V., Yefymchykov O.M., Korkin O.Y., Goncharuk A.A.
Solving the problem of generating stable and consistent estimates of the correlation matrix of observations by the dynamic regularization method

We analyze the consistency of stable estimates of a correlation matrix of observations at their static and dynamic regularization. We prove the advantage of the dynamic regularization method with the optimal parameter in the context of resolving the contradiction “computational stability–consistency” of sample estimates of a correlation matrix of observations. We obtain the algorithm for computation of the optimal parameter of dynamic regularization that do not use predicted data and do not require additional computational costs. © 2021, Springer Science+Business Media, LLC, part of Springer Nature.

Keywords: consistency, convergence, estimate, regularization, stability, Matrix algebra, Computational costs, Computational stability, Correlation matrix, Dynamic regularizations, Optimal parameter, Parameter estimation


Cite:
Skachkov V.V., Chepkii V.V., Yefymchykov O.M., Korkin O.Y., Goncharuk A.A. (2021). Solving the problem of generating stable and consistent estimates of the correlation matrix of observations by the dynamic regularization method. Cybernetics and Systems Analysis, 57 (1), 94–103. doi: https://doi.org/10.1007/s10559-021-00331-3 http://jnas.nbuv.gov.ua/article/UJRN-0001199859 [In Russian].


 

Інститут інформаційних технологій НБУВ


+38 (044) 525-36-24
Голосіївський просп., 3, к. 209
м. Київ, 03039, Україна