A Randomized Algorithm for Parsimonious Model Identification

Burak Yilmaz, Korkut Bekiroglu, Constantino Manuel Lagoa, Mario Sznaier

Research output: Contribution to journalArticle

4 Citations (Scopus)

Abstract

Identifying parsimonious models is generically a 'hard' nonconvex problem. Available approaches typically rely on relaxations such as Group Lasso or nuclear norm minimization. Moreover, incorporating stability and model order constraints into the formalism in such methods entails a substantial increase in computational complexity. Motivated by these challenges, in this paper we present algorithms for parsimonious linear time invariant system identification aimed at identifying low-complexity models which i) incorporate a priori knowledge on the system (e.g., stability), ii) allow for data with missing/nonuniform measurements, and iii) are able to use data obtained from several runs of the system with different unknown initial conditions. The randomized algorithms proposed are based on the concept of atomic norm and provide a numerically efficient way to identify sparse models from large amounts of noisy data.

Original languageEnglish (US)
Article number7970196
Pages (from-to)532-539
Number of pages8
JournalIEEE Transactions on Automatic Control
Volume63
Issue number2
DOIs
StatePublished - Feb 1 2018

Fingerprint

Identification (control systems)
System stability
Computational complexity

All Science Journal Classification (ASJC) codes

  • Control and Systems Engineering
  • Computer Science Applications
  • Electrical and Electronic Engineering

Cite this

Yilmaz, Burak ; Bekiroglu, Korkut ; Lagoa, Constantino Manuel ; Sznaier, Mario. / A Randomized Algorithm for Parsimonious Model Identification. In: IEEE Transactions on Automatic Control. 2018 ; Vol. 63, No. 2. pp. 532-539.
@article{c45774ed5f4e4222999fe72642cd6adb,
title = "A Randomized Algorithm for Parsimonious Model Identification",
abstract = "Identifying parsimonious models is generically a 'hard' nonconvex problem. Available approaches typically rely on relaxations such as Group Lasso or nuclear norm minimization. Moreover, incorporating stability and model order constraints into the formalism in such methods entails a substantial increase in computational complexity. Motivated by these challenges, in this paper we present algorithms for parsimonious linear time invariant system identification aimed at identifying low-complexity models which i) incorporate a priori knowledge on the system (e.g., stability), ii) allow for data with missing/nonuniform measurements, and iii) are able to use data obtained from several runs of the system with different unknown initial conditions. The randomized algorithms proposed are based on the concept of atomic norm and provide a numerically efficient way to identify sparse models from large amounts of noisy data.",
author = "Burak Yilmaz and Korkut Bekiroglu and Lagoa, {Constantino Manuel} and Mario Sznaier",
year = "2018",
month = "2",
day = "1",
doi = "10.1109/TAC.2017.2723959",
language = "English (US)",
volume = "63",
pages = "532--539",
journal = "IEEE Transactions on Automatic Control",
issn = "0018-9286",
publisher = "Institute of Electrical and Electronics Engineers Inc.",
number = "2",

}

A Randomized Algorithm for Parsimonious Model Identification. / Yilmaz, Burak; Bekiroglu, Korkut; Lagoa, Constantino Manuel; Sznaier, Mario.

In: IEEE Transactions on Automatic Control, Vol. 63, No. 2, 7970196, 01.02.2018, p. 532-539.

Research output: Contribution to journalArticle

TY - JOUR

T1 - A Randomized Algorithm for Parsimonious Model Identification

AU - Yilmaz, Burak

AU - Bekiroglu, Korkut

AU - Lagoa, Constantino Manuel

AU - Sznaier, Mario

PY - 2018/2/1

Y1 - 2018/2/1

N2 - Identifying parsimonious models is generically a 'hard' nonconvex problem. Available approaches typically rely on relaxations such as Group Lasso or nuclear norm minimization. Moreover, incorporating stability and model order constraints into the formalism in such methods entails a substantial increase in computational complexity. Motivated by these challenges, in this paper we present algorithms for parsimonious linear time invariant system identification aimed at identifying low-complexity models which i) incorporate a priori knowledge on the system (e.g., stability), ii) allow for data with missing/nonuniform measurements, and iii) are able to use data obtained from several runs of the system with different unknown initial conditions. The randomized algorithms proposed are based on the concept of atomic norm and provide a numerically efficient way to identify sparse models from large amounts of noisy data.

AB - Identifying parsimonious models is generically a 'hard' nonconvex problem. Available approaches typically rely on relaxations such as Group Lasso or nuclear norm minimization. Moreover, incorporating stability and model order constraints into the formalism in such methods entails a substantial increase in computational complexity. Motivated by these challenges, in this paper we present algorithms for parsimonious linear time invariant system identification aimed at identifying low-complexity models which i) incorporate a priori knowledge on the system (e.g., stability), ii) allow for data with missing/nonuniform measurements, and iii) are able to use data obtained from several runs of the system with different unknown initial conditions. The randomized algorithms proposed are based on the concept of atomic norm and provide a numerically efficient way to identify sparse models from large amounts of noisy data.

UR - http://www.scopus.com/inward/record.url?scp=85023208951&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85023208951&partnerID=8YFLogxK

U2 - 10.1109/TAC.2017.2723959

DO - 10.1109/TAC.2017.2723959

M3 - Article

AN - SCOPUS:85023208951

VL - 63

SP - 532

EP - 539

JO - IEEE Transactions on Automatic Control

JF - IEEE Transactions on Automatic Control

SN - 0018-9286

IS - 2

M1 - 7970196

ER -