Positive and unlabeled multi-graph learning

Jia Wu, Shirui Pan, Xingquan Zhu, Chengqi Zhang, Xindong Wu

Research output: Contribution to journalArticleResearchpeer-review

Abstract

In this paper, we advance graph classification to handle multi-graph learning for complicated objects, where each object is represented as a bag of graphs and the label is only available to each bag but not individual graphs. In addition, when training classifiers, users are only given a handful of positive bags and many unlabeled bags, and the learning objective is to train models to classify previously unseen graph bags with maximum accuracy. To achieve the goal, we propose a positive and unlabeled multi-graph learning (puMGL) framework to first select informative subgraphs to convert graphs into a feature space. To utilize unlabeled bags for learning, puMGL assigns a confidence weight to each bag and dynamically adjusts its weight value to select 'reliable negative bags.' A number of representative graphs, selected from positive bags and identified reliable negative graph bags, form a 'margin graph pool' which serves as the base for deriving subgraph patterns, training graph classifiers, and further updating the bag weight values. A closed-loop iterative process helps discover optimal subgraphs from positive and unlabeled graph bags for learning. Experimental comparisons demonstrate the performance of puMGL for classifying real-world complicated objects.

Original languageEnglish
Article number7439802
Pages (from-to)818-829
Number of pages12
JournalIEEE Transactions on Cybernetics
Volume47
Issue number4
DOIs
Publication statusPublished - Apr 2017
Externally publishedYes

Keywords

  • Classification
  • features
  • Graph
  • multi-instance (MI)
  • positive and unlabeled (PU) learning
  • subgraph

Cite this

Wu, Jia ; Pan, Shirui ; Zhu, Xingquan ; Zhang, Chengqi ; Wu, Xindong. / Positive and unlabeled multi-graph learning. In: IEEE Transactions on Cybernetics. 2017 ; Vol. 47, No. 4. pp. 818-829.
@article{a962e275ddcc4a6f82d0ddf3e99c2466,
title = "Positive and unlabeled multi-graph learning",
abstract = "In this paper, we advance graph classification to handle multi-graph learning for complicated objects, where each object is represented as a bag of graphs and the label is only available to each bag but not individual graphs. In addition, when training classifiers, users are only given a handful of positive bags and many unlabeled bags, and the learning objective is to train models to classify previously unseen graph bags with maximum accuracy. To achieve the goal, we propose a positive and unlabeled multi-graph learning (puMGL) framework to first select informative subgraphs to convert graphs into a feature space. To utilize unlabeled bags for learning, puMGL assigns a confidence weight to each bag and dynamically adjusts its weight value to select 'reliable negative bags.' A number of representative graphs, selected from positive bags and identified reliable negative graph bags, form a 'margin graph pool' which serves as the base for deriving subgraph patterns, training graph classifiers, and further updating the bag weight values. A closed-loop iterative process helps discover optimal subgraphs from positive and unlabeled graph bags for learning. Experimental comparisons demonstrate the performance of puMGL for classifying real-world complicated objects.",
keywords = "Classification, features, Graph, multi-instance (MI), positive and unlabeled (PU) learning, subgraph",
author = "Jia Wu and Shirui Pan and Xingquan Zhu and Chengqi Zhang and Xindong Wu",
year = "2017",
month = "4",
doi = "10.1109/TCYB.2016.2527239",
language = "English",
volume = "47",
pages = "818--829",
journal = "IEEE Transactions on Cybernetics",
issn = "2168-2267",
number = "4",

}

Positive and unlabeled multi-graph learning. / Wu, Jia; Pan, Shirui; Zhu, Xingquan; Zhang, Chengqi; Wu, Xindong.

In: IEEE Transactions on Cybernetics, Vol. 47, No. 4, 7439802, 04.2017, p. 818-829.

Research output: Contribution to journalArticleResearchpeer-review

TY - JOUR

T1 - Positive and unlabeled multi-graph learning

AU - Wu, Jia

AU - Pan, Shirui

AU - Zhu, Xingquan

AU - Zhang, Chengqi

AU - Wu, Xindong

PY - 2017/4

Y1 - 2017/4

N2 - In this paper, we advance graph classification to handle multi-graph learning for complicated objects, where each object is represented as a bag of graphs and the label is only available to each bag but not individual graphs. In addition, when training classifiers, users are only given a handful of positive bags and many unlabeled bags, and the learning objective is to train models to classify previously unseen graph bags with maximum accuracy. To achieve the goal, we propose a positive and unlabeled multi-graph learning (puMGL) framework to first select informative subgraphs to convert graphs into a feature space. To utilize unlabeled bags for learning, puMGL assigns a confidence weight to each bag and dynamically adjusts its weight value to select 'reliable negative bags.' A number of representative graphs, selected from positive bags and identified reliable negative graph bags, form a 'margin graph pool' which serves as the base for deriving subgraph patterns, training graph classifiers, and further updating the bag weight values. A closed-loop iterative process helps discover optimal subgraphs from positive and unlabeled graph bags for learning. Experimental comparisons demonstrate the performance of puMGL for classifying real-world complicated objects.

AB - In this paper, we advance graph classification to handle multi-graph learning for complicated objects, where each object is represented as a bag of graphs and the label is only available to each bag but not individual graphs. In addition, when training classifiers, users are only given a handful of positive bags and many unlabeled bags, and the learning objective is to train models to classify previously unseen graph bags with maximum accuracy. To achieve the goal, we propose a positive and unlabeled multi-graph learning (puMGL) framework to first select informative subgraphs to convert graphs into a feature space. To utilize unlabeled bags for learning, puMGL assigns a confidence weight to each bag and dynamically adjusts its weight value to select 'reliable negative bags.' A number of representative graphs, selected from positive bags and identified reliable negative graph bags, form a 'margin graph pool' which serves as the base for deriving subgraph patterns, training graph classifiers, and further updating the bag weight values. A closed-loop iterative process helps discover optimal subgraphs from positive and unlabeled graph bags for learning. Experimental comparisons demonstrate the performance of puMGL for classifying real-world complicated objects.

KW - Classification

KW - features

KW - Graph

KW - multi-instance (MI)

KW - positive and unlabeled (PU) learning

KW - subgraph

UR - http://www.scopus.com/inward/record.url?scp=84961933000&partnerID=8YFLogxK

U2 - 10.1109/TCYB.2016.2527239

DO - 10.1109/TCYB.2016.2527239

M3 - Article

VL - 47

SP - 818

EP - 829

JO - IEEE Transactions on Cybernetics

JF - IEEE Transactions on Cybernetics

SN - 2168-2267

IS - 4

M1 - 7439802

ER -