A neural network model for sequential multitask pattern recognition problems

Hitoshi Nishikawa, Seiichi Ozawa, Asim Roy

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

Abstract

In this paper, we propose a new multitask learning (MTL) model which can learn a series of multi-class pattern recognition prob- lems stably. The knowledge transfer in the proposed MTL model is implemented by the following mechanisms: (1) transfer by sharing the internal representation of RBFs and (2) transfer of the information on class subregions from the related tasks. The proposed model can detect task changes on its own based on the output errors even though no task information is given by the environment. It also learn training samples of different tasks that are given one after another. In the experiments, the recognition performance is evaluated for the eight MTPR problems which are defined from the four UCI data sets. The experimental results demonstrate that the proposed MTL model outperforms a single-task learning model in terms of the final classification accuracy. Furthermore, we show that the transfer of class subregion contributes to enhancing the generalization performance of a new task with less training samples.

Original languageEnglish (US)
Title of host publicationLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Pages821-828
Number of pages8
Volume5506 LNCS
EditionPART 1
DOIs
StatePublished - 2009
Event15th International Conference on Neuro-Information Processing, ICONIP 2008 - Auckland, New Zealand
Duration: Nov 25 2008Nov 28 2008

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
NumberPART 1
Volume5506 LNCS
ISSN (Print)03029743
ISSN (Electronic)16113349

Other

Other15th International Conference on Neuro-Information Processing, ICONIP 2008
CountryNew Zealand
CityAuckland
Period11/25/0811/28/08

Fingerprint

Neural Network Model
Pattern Recognition
Pattern recognition
Multi-task Learning
Neural networks
Training Samples
Knowledge Transfer
Model
Multi-class
Sharing
Internal
Series
Output
Experimental Results
Demonstrate
Experiment
Experiments
Class

ASJC Scopus subject areas

  • Computer Science(all)
  • Theoretical Computer Science

Cite this

Nishikawa, H., Ozawa, S., & Roy, A. (2009). A neural network model for sequential multitask pattern recognition problems. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) (PART 1 ed., Vol. 5506 LNCS, pp. 821-828). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 5506 LNCS, No. PART 1). https://doi.org/10.1007/978-3-642-02490-0_100

A neural network model for sequential multitask pattern recognition problems. / Nishikawa, Hitoshi; Ozawa, Seiichi; Roy, Asim.

Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). Vol. 5506 LNCS PART 1. ed. 2009. p. 821-828 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 5506 LNCS, No. PART 1).

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

Nishikawa, H, Ozawa, S & Roy, A 2009, A neural network model for sequential multitask pattern recognition problems. in Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). PART 1 edn, vol. 5506 LNCS, Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), no. PART 1, vol. 5506 LNCS, pp. 821-828, 15th International Conference on Neuro-Information Processing, ICONIP 2008, Auckland, New Zealand, 11/25/08. https://doi.org/10.1007/978-3-642-02490-0_100
Nishikawa H, Ozawa S, Roy A. A neural network model for sequential multitask pattern recognition problems. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). PART 1 ed. Vol. 5506 LNCS. 2009. p. 821-828. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); PART 1). https://doi.org/10.1007/978-3-642-02490-0_100
Nishikawa, Hitoshi ; Ozawa, Seiichi ; Roy, Asim. / A neural network model for sequential multitask pattern recognition problems. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). Vol. 5506 LNCS PART 1. ed. 2009. pp. 821-828 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); PART 1).
@inproceedings{dc819bad6f934522826c5f5015587580,
title = "A neural network model for sequential multitask pattern recognition problems",
abstract = "In this paper, we propose a new multitask learning (MTL) model which can learn a series of multi-class pattern recognition prob- lems stably. The knowledge transfer in the proposed MTL model is implemented by the following mechanisms: (1) transfer by sharing the internal representation of RBFs and (2) transfer of the information on class subregions from the related tasks. The proposed model can detect task changes on its own based on the output errors even though no task information is given by the environment. It also learn training samples of different tasks that are given one after another. In the experiments, the recognition performance is evaluated for the eight MTPR problems which are defined from the four UCI data sets. The experimental results demonstrate that the proposed MTL model outperforms a single-task learning model in terms of the final classification accuracy. Furthermore, we show that the transfer of class subregion contributes to enhancing the generalization performance of a new task with less training samples.",
author = "Hitoshi Nishikawa and Seiichi Ozawa and Asim Roy",
year = "2009",
doi = "10.1007/978-3-642-02490-0_100",
language = "English (US)",
isbn = "3642024890",
volume = "5506 LNCS",
series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",
number = "PART 1",
pages = "821--828",
booktitle = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",
edition = "PART 1",

}

TY - GEN

T1 - A neural network model for sequential multitask pattern recognition problems

AU - Nishikawa, Hitoshi

AU - Ozawa, Seiichi

AU - Roy, Asim

PY - 2009

Y1 - 2009

N2 - In this paper, we propose a new multitask learning (MTL) model which can learn a series of multi-class pattern recognition prob- lems stably. The knowledge transfer in the proposed MTL model is implemented by the following mechanisms: (1) transfer by sharing the internal representation of RBFs and (2) transfer of the information on class subregions from the related tasks. The proposed model can detect task changes on its own based on the output errors even though no task information is given by the environment. It also learn training samples of different tasks that are given one after another. In the experiments, the recognition performance is evaluated for the eight MTPR problems which are defined from the four UCI data sets. The experimental results demonstrate that the proposed MTL model outperforms a single-task learning model in terms of the final classification accuracy. Furthermore, we show that the transfer of class subregion contributes to enhancing the generalization performance of a new task with less training samples.

AB - In this paper, we propose a new multitask learning (MTL) model which can learn a series of multi-class pattern recognition prob- lems stably. The knowledge transfer in the proposed MTL model is implemented by the following mechanisms: (1) transfer by sharing the internal representation of RBFs and (2) transfer of the information on class subregions from the related tasks. The proposed model can detect task changes on its own based on the output errors even though no task information is given by the environment. It also learn training samples of different tasks that are given one after another. In the experiments, the recognition performance is evaluated for the eight MTPR problems which are defined from the four UCI data sets. The experimental results demonstrate that the proposed MTL model outperforms a single-task learning model in terms of the final classification accuracy. Furthermore, we show that the transfer of class subregion contributes to enhancing the generalization performance of a new task with less training samples.

UR - http://www.scopus.com/inward/record.url?scp=70349135509&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=70349135509&partnerID=8YFLogxK

U2 - 10.1007/978-3-642-02490-0_100

DO - 10.1007/978-3-642-02490-0_100

M3 - Conference contribution

SN - 3642024890

SN - 9783642024894

VL - 5506 LNCS

T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

SP - 821

EP - 828

BT - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

ER -