NARX Identification using Derivative-Based Regularized Neural Networks

L.H. Peeters; G.I. Beintema; M. Forgione; M. Schoukens

doi:10.1109/CDC51059.2022.9992559

NARX Identification using Derivative-Based Regularized Neural Networks

L.H. Peeters, G.I. Beintema, M. Forgione, M. Schoukens

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › Academic › peer-review

1 Citation (Scopus)

112 Downloads (Pure)

Abstract

This work presents a novel regularization method for the identification of Nonlinear Autoregressive eXogenous (NARX) models. The regularization method promotes the exponential decay of the influence of past input samples on the current model output. This is done by penalizing the sensitivity of the NARX model simulated output with respect to the past inputs. This promotes the stability of the estimated models and improves the obtained model quality. The effectiveness of the approach is demonstrated through a simulation example, where a neural network NARX model is identified with this novel method. Moreover, it is shown that the proposed regularization approach improves the model accuracy in terms of simulation error performance compared to that of other regularization methods and model classes.

Original language	English
Title of host publication	61st IEEE Conference on Decision and Control
Publisher	Institute of Electrical and Electronics Engineers
Pages	1515-1520
Number of pages	6
ISBN (Electronic)	978-1-6654-6761-2
DOIs	https://doi.org/10.1109/CDC51059.2022.9992559
Publication status	Published - 10 Jan 2023
Event	61st IEEE Conference on Decision and Control, CDC 2022 - The Marriott Cancún Collection, Cancun, Mexico Duration: 6 Dec 2022 → 9 Dec 2022 Conference number: 61 https://cdc2022.ieeecss.org/

Conference

Conference	61st IEEE Conference on Decision and Control, CDC 2022
Abbreviated title	CDC 2022
Country/Territory	Mexico
City	Cancun
Period	6/12/22 → 9/12/22
Internet address	https://cdc2022.ieeecss.org/

Access to Document

10.1109/CDC51059.2022.9992559

2204.05892v2Final author version , 538 KB

Cite this

@inproceedings{8dea90f86e6e438293180c80260030a6,

title = "NARX Identification using Derivative-Based Regularized Neural Networks",

abstract = "This work presents a novel regularization method for the identification of Nonlinear Autoregressive eXogenous (NARX) models. The regularization method promotes the exponential decay of the influence of past input samples on the current model output. This is done by penalizing the sensitivity of the NARX model simulated output with respect to the past inputs. This promotes the stability of the estimated models and improves the obtained model quality. The effectiveness of the approach is demonstrated through a simulation example, where a neural network NARX model is identified with this novel method. Moreover, it is shown that the proposed regularization approach improves the model accuracy in terms of simulation error performance compared to that of other regularization methods and model classes. ",

keywords = "eess.SY, cs.LG, cs.SY",

author = "L.H. Peeters and G.I. Beintema and M. Forgione and M. Schoukens",

year = "2023",

month = jan,

day = "10",

doi = "10.1109/CDC51059.2022.9992559",

language = "English",

pages = "1515--1520",

booktitle = "61st IEEE Conference on Decision and Control",

publisher = "Institute of Electrical and Electronics Engineers",

address = "United States",

note = "61st IEEE Conference on Decision and Control, CDC 2022, CDC 2022 ; Conference date: 06-12-2022 Through 09-12-2022",

url = "https://cdc2022.ieeecss.org/",

}

Peeters, LH, Beintema, GI, Forgione, M & Schoukens, M 2023, NARX Identification using Derivative-Based Regularized Neural Networks. in 61st IEEE Conference on Decision and Control ., 9992559, Institute of Electrical and Electronics Engineers, pp. 1515-1520, 61st IEEE Conference on Decision and Control, CDC 2022, Cancun, Mexico, 6/12/22. https://doi.org/10.1109/CDC51059.2022.9992559

TY - GEN

T1 - NARX Identification using Derivative-Based Regularized Neural Networks

AU - Peeters, L.H.

AU - Beintema, G.I.

AU - Forgione, M.

AU - Schoukens, M.

N1 - Conference code: 61

PY - 2023/1/10

Y1 - 2023/1/10

N2 - This work presents a novel regularization method for the identification of Nonlinear Autoregressive eXogenous (NARX) models. The regularization method promotes the exponential decay of the influence of past input samples on the current model output. This is done by penalizing the sensitivity of the NARX model simulated output with respect to the past inputs. This promotes the stability of the estimated models and improves the obtained model quality. The effectiveness of the approach is demonstrated through a simulation example, where a neural network NARX model is identified with this novel method. Moreover, it is shown that the proposed regularization approach improves the model accuracy in terms of simulation error performance compared to that of other regularization methods and model classes.

AB - This work presents a novel regularization method for the identification of Nonlinear Autoregressive eXogenous (NARX) models. The regularization method promotes the exponential decay of the influence of past input samples on the current model output. This is done by penalizing the sensitivity of the NARX model simulated output with respect to the past inputs. This promotes the stability of the estimated models and improves the obtained model quality. The effectiveness of the approach is demonstrated through a simulation example, where a neural network NARX model is identified with this novel method. Moreover, it is shown that the proposed regularization approach improves the model accuracy in terms of simulation error performance compared to that of other regularization methods and model classes.

KW - eess.SY

KW - cs.LG

KW - cs.SY

UR - http://www.scopus.com/inward/record.url?scp=85147014805&partnerID=8YFLogxK

U2 - 10.1109/CDC51059.2022.9992559

DO - 10.1109/CDC51059.2022.9992559

M3 - Conference contribution

SP - 1515

EP - 1520

BT - 61st IEEE Conference on Decision and Control

PB - Institute of Electrical and Electronics Engineers

T2 - 61st IEEE Conference on Decision and Control, CDC 2022

Y2 - 6 December 2022 through 9 December 2022

ER -

NARX Identification using Derivative-Based Regularized Neural Networks

Abstract

Conference

Access to Document

Other files and links

Fingerprint

Cite this