Towards D-optimal input design for finite-sample system identification

Sándor Kolumbán, Balázs Csanád Csáji

Onderzoeksoutput: Bijdrage aan tijdschriftCongresartikelpeer review

1 Citaat (Scopus)

Samenvatting

Finite-sample system identification methods provide statistical inference, typically in the form of confidence regions, with rigorous non-asymptotic guarantees under minimal distributional assumptions. Data Perturbation (DP) methods constitute an important class of such algorithms, which includes, for example, Sign-Perturbed Sums (SPS) as a special case. Here we study a natural input design problem for DP methods in linear regression models, where we want to select the regressors in a way that the expected volume of the resulting confidence regions are minimized. We suggest a general approach to this problem and analyze it for the fundamental building blocks of all DP confidence regions, namely, for ellipsoids having confidence probability exactly 1/2. We also present experiments supporting that minimizing the expected volumes of such ellipsoids significantly reduces the average sizes of the constructed DP confidence regions.

Originele taal-2Engels
Pagina's (van-tot)215-220
Aantal pagina's6
TijdschriftIFAC-PapersOnLine
Volume51
Nummer van het tijdschrift15
DOI's
StatusGepubliceerd - 8 okt. 2018
Evenement18th IFAC Symposium on System Identification (SYSID 2018) - Stockholm, Zweden
Duur: 9 jul. 201811 jul. 2018

Vingerafdruk

Duik in de onderzoeksthema's van 'Towards D-optimal input design for finite-sample system identification'. Samen vormen ze een unieke vingerafdruk.

Citeer dit