Dynamic sampling approach to training neural networks for multiclass imbalance classification

Minlong LIN; Ke TANG; Xin YAO

doi:10.1109/TNNLS.2012.2228231

Dynamic sampling approach to training neural networks for multiclass imbalance classification

Minlong LIN, Ke TANG, Xin YAO

Research output: Journal Publications › Journal Article (refereed) › peer-review

148 Citations (Scopus)

Abstract

Class imbalance learning tackles supervised learning problems where some classes have significantly more examples than others. Most of the existing research focused only on binary-class cases. In this paper, we study multiclass imbalance problems and propose a dynamic sampling method (DyS) for multilayer perceptrons (MLP). In DyS, for each epoch of the training process, every example is fed to the current MLP and then the probability of it being selected for training the MLP is estimated. DyS dynamically selects informative data to train the MLP. In order to evaluate DyS and understand its strength and weakness, comprehensive experimental studies have been carried out. Results on 20 multiclass imbalanced data sets show that DyS can outperform the compared methods, including pre-sample methods, active learning methods, cost-sensitive methods, and boosting-type methods. © 2012 IEEE.

Original language	English
Article number	6449324
Pages (from-to)	647-660
Number of pages	14
Journal	IEEE Transactions on Neural Networks and Learning Systems
Volume	24
Issue number	4
Early online date	4 Feb 2013
DOIs	https://doi.org/10.1109/TNNLS.2012.2228231
Publication status	Published - Apr 2013
Externally published	Yes

Funding

This work was supported in part by the 973 Program of China under Grant 2011CB707006, the National Natural Science Foundation of China under Grant 61175065, Grant U0835002, and Grant 61028009, the National Natural Science Foundation of Anhui Province under Grant 1108085J16, and the European Union Seventh Framework Programme under Grant 247619 and Grant 270428.

Keywords

Cost-sensitive learning
dynamic sampling
multiclass imbalance learning
multilayer perceptrons

Access to Document

10.1109/TNNLS.2012.2228231

Cite this

@article{de5a4d3fda954a219e489059d687ceb6,

title = "Dynamic sampling approach to training neural networks for multiclass imbalance classification",

abstract = "Class imbalance learning tackles supervised learning problems where some classes have significantly more examples than others. Most of the existing research focused only on binary-class cases. In this paper, we study multiclass imbalance problems and propose a dynamic sampling method (DyS) for multilayer perceptrons (MLP). In DyS, for each epoch of the training process, every example is fed to the current MLP and then the probability of it being selected for training the MLP is estimated. DyS dynamically selects informative data to train the MLP. In order to evaluate DyS and understand its strength and weakness, comprehensive experimental studies have been carried out. Results on 20 multiclass imbalanced data sets show that DyS can outperform the compared methods, including pre-sample methods, active learning methods, cost-sensitive methods, and boosting-type methods. {\textcopyright} 2012 IEEE.",

keywords = "Cost-sensitive learning, dynamic sampling, multiclass imbalance learning, multilayer perceptrons",

author = "Minlong LIN and Ke TANG and Xin YAO",

year = "2013",

month = apr,

doi = "10.1109/TNNLS.2012.2228231",

language = "English",

volume = "24",

pages = "647--660",

journal = "IEEE Transactions on Neural Networks and Learning Systems",

issn = "2162-237X",

publisher = "IEEE Computational Intelligence Society",

number = "4",

}

TY - JOUR

T1 - Dynamic sampling approach to training neural networks for multiclass imbalance classification

AU - LIN, Minlong

AU - TANG, Ke

AU - YAO, Xin

PY - 2013/4

Y1 - 2013/4

N2 - Class imbalance learning tackles supervised learning problems where some classes have significantly more examples than others. Most of the existing research focused only on binary-class cases. In this paper, we study multiclass imbalance problems and propose a dynamic sampling method (DyS) for multilayer perceptrons (MLP). In DyS, for each epoch of the training process, every example is fed to the current MLP and then the probability of it being selected for training the MLP is estimated. DyS dynamically selects informative data to train the MLP. In order to evaluate DyS and understand its strength and weakness, comprehensive experimental studies have been carried out. Results on 20 multiclass imbalanced data sets show that DyS can outperform the compared methods, including pre-sample methods, active learning methods, cost-sensitive methods, and boosting-type methods. © 2012 IEEE.

AB - Class imbalance learning tackles supervised learning problems where some classes have significantly more examples than others. Most of the existing research focused only on binary-class cases. In this paper, we study multiclass imbalance problems and propose a dynamic sampling method (DyS) for multilayer perceptrons (MLP). In DyS, for each epoch of the training process, every example is fed to the current MLP and then the probability of it being selected for training the MLP is estimated. DyS dynamically selects informative data to train the MLP. In order to evaluate DyS and understand its strength and weakness, comprehensive experimental studies have been carried out. Results on 20 multiclass imbalanced data sets show that DyS can outperform the compared methods, including pre-sample methods, active learning methods, cost-sensitive methods, and boosting-type methods. © 2012 IEEE.

KW - Cost-sensitive learning

KW - dynamic sampling

KW - multiclass imbalance learning

KW - multilayer perceptrons

UR - http://www.scopus.com/inward/record.url?scp=84875898112&partnerID=8YFLogxK

U2 - 10.1109/TNNLS.2012.2228231

DO - 10.1109/TNNLS.2012.2228231

M3 - Journal Article (refereed)

SN - 2162-237X

VL - 24

SP - 647

EP - 660

JO - IEEE Transactions on Neural Networks and Learning Systems

JF - IEEE Transactions on Neural Networks and Learning Systems

IS - 4

M1 - 6449324

ER -

Dynamic sampling approach to training neural networks for multiclass imbalance classification

Abstract

Funding

Keywords

Access to Document

Other files and links

Fingerprint

Cite this