Filtering noise in regression problems using a multiobjective leaning algorithm

D. A G Vieira, X. L. Travassos, Vasile Palade, R. R. Saldanha

Research output: Chapter in Book/Report/Conference proceedingConference proceeding

Abstract

This paper applies a Neural Networks (NN) multiobjective learning algorithm called the Minimum Gradient Method (MGM) to filter noise in regression problems. This method is based on the concept that the learning is a bi-objective problem aiming at minimizing the empirical risk (training error) and the function complexity. The complexity is modeled as the norm of the network output gradient. After training, the NN behaves as an adaptive filter which minimizes the cross-validation error. The NN trained with this method can be used to pre-process the data and help reduce the signal-to-noise ratio (SNR). Some results are presented and they show the effectiveness of the proposed approach.

Original languageEnglish
Title of host publicationProceedings - 20th IEEE International Conference on Tools with Artificial Intelligence, ICTAI'08
PublisherIEEE
Pages452-456
Number of pages5
Volume2
ISBN (Print)9780769534404
DOIs
Publication statusPublished - 2008
Externally publishedYes
Event20th IEEE International Conference on Tools with Artificial Intelligence, ICTAI'08 - Dayton, OH, United States
Duration: 3 Nov 20085 Nov 2008

Conference

Conference20th IEEE International Conference on Tools with Artificial Intelligence, ICTAI'08
CountryUnited States
CityDayton, OH
Period3/11/085/11/08

Fingerprint

Neural networks
Gradient methods
Adaptive filters
Learning algorithms
Signal to noise ratio

Keywords

  • Inverse problems
  • Multiobjective training algorithms
  • Neural networks
  • Noise
  • Regression problems
  • Regularization methods

ASJC Scopus subject areas

  • Software
  • Artificial Intelligence
  • Computer Science Applications

Cite this

Vieira, D. A. G., Travassos, X. L., Palade, V., & Saldanha, R. R. (2008). Filtering noise in regression problems using a multiobjective leaning algorithm. In Proceedings - 20th IEEE International Conference on Tools with Artificial Intelligence, ICTAI'08 (Vol. 2, pp. 452-456). [4669808] IEEE. https://doi.org/10.1109/ICTAI.2008.17

Filtering noise in regression problems using a multiobjective leaning algorithm. / Vieira, D. A G; Travassos, X. L.; Palade, Vasile; Saldanha, R. R.

Proceedings - 20th IEEE International Conference on Tools with Artificial Intelligence, ICTAI'08. Vol. 2 IEEE, 2008. p. 452-456 4669808.

Research output: Chapter in Book/Report/Conference proceedingConference proceeding

Vieira, DAG, Travassos, XL, Palade, V & Saldanha, RR 2008, Filtering noise in regression problems using a multiobjective leaning algorithm. in Proceedings - 20th IEEE International Conference on Tools with Artificial Intelligence, ICTAI'08. vol. 2, 4669808, IEEE, pp. 452-456, 20th IEEE International Conference on Tools with Artificial Intelligence, ICTAI'08, Dayton, OH, United States, 3/11/08. https://doi.org/10.1109/ICTAI.2008.17
Vieira DAG, Travassos XL, Palade V, Saldanha RR. Filtering noise in regression problems using a multiobjective leaning algorithm. In Proceedings - 20th IEEE International Conference on Tools with Artificial Intelligence, ICTAI'08. Vol. 2. IEEE. 2008. p. 452-456. 4669808 https://doi.org/10.1109/ICTAI.2008.17
Vieira, D. A G ; Travassos, X. L. ; Palade, Vasile ; Saldanha, R. R. / Filtering noise in regression problems using a multiobjective leaning algorithm. Proceedings - 20th IEEE International Conference on Tools with Artificial Intelligence, ICTAI'08. Vol. 2 IEEE, 2008. pp. 452-456
@inproceedings{ac1ed96c3b0e4c549d0c9e221af52763,
title = "Filtering noise in regression problems using a multiobjective leaning algorithm",
abstract = "This paper applies a Neural Networks (NN) multiobjective learning algorithm called the Minimum Gradient Method (MGM) to filter noise in regression problems. This method is based on the concept that the learning is a bi-objective problem aiming at minimizing the empirical risk (training error) and the function complexity. The complexity is modeled as the norm of the network output gradient. After training, the NN behaves as an adaptive filter which minimizes the cross-validation error. The NN trained with this method can be used to pre-process the data and help reduce the signal-to-noise ratio (SNR). Some results are presented and they show the effectiveness of the proposed approach.",
keywords = "Inverse problems, Multiobjective training algorithms, Neural networks, Noise, Regression problems, Regularization methods",
author = "Vieira, {D. A G} and Travassos, {X. L.} and Vasile Palade and Saldanha, {R. R.}",
year = "2008",
doi = "10.1109/ICTAI.2008.17",
language = "English",
isbn = "9780769534404",
volume = "2",
pages = "452--456",
booktitle = "Proceedings - 20th IEEE International Conference on Tools with Artificial Intelligence, ICTAI'08",
publisher = "IEEE",

}

TY - GEN

T1 - Filtering noise in regression problems using a multiobjective leaning algorithm

AU - Vieira, D. A G

AU - Travassos, X. L.

AU - Palade, Vasile

AU - Saldanha, R. R.

PY - 2008

Y1 - 2008

N2 - This paper applies a Neural Networks (NN) multiobjective learning algorithm called the Minimum Gradient Method (MGM) to filter noise in regression problems. This method is based on the concept that the learning is a bi-objective problem aiming at minimizing the empirical risk (training error) and the function complexity. The complexity is modeled as the norm of the network output gradient. After training, the NN behaves as an adaptive filter which minimizes the cross-validation error. The NN trained with this method can be used to pre-process the data and help reduce the signal-to-noise ratio (SNR). Some results are presented and they show the effectiveness of the proposed approach.

AB - This paper applies a Neural Networks (NN) multiobjective learning algorithm called the Minimum Gradient Method (MGM) to filter noise in regression problems. This method is based on the concept that the learning is a bi-objective problem aiming at minimizing the empirical risk (training error) and the function complexity. The complexity is modeled as the norm of the network output gradient. After training, the NN behaves as an adaptive filter which minimizes the cross-validation error. The NN trained with this method can be used to pre-process the data and help reduce the signal-to-noise ratio (SNR). Some results are presented and they show the effectiveness of the proposed approach.

KW - Inverse problems

KW - Multiobjective training algorithms

KW - Neural networks

KW - Noise

KW - Regression problems

KW - Regularization methods

UR - http://www.scopus.com/inward/record.url?scp=57649221170&partnerID=8YFLogxK

U2 - 10.1109/ICTAI.2008.17

DO - 10.1109/ICTAI.2008.17

M3 - Conference proceeding

SN - 9780769534404

VL - 2

SP - 452

EP - 456

BT - Proceedings - 20th IEEE International Conference on Tools with Artificial Intelligence, ICTAI'08

PB - IEEE

ER -