Joint structure feature exploration and regularization for multi-task graph classification

Shirui Pan, Jia Wu, Xingquan Zhu, Chengqi Zhang, Philip S. Yu

Research output: Contribution to journalArticleResearchpeer-review

Abstract

Graph classification aims to learn models to classify structure data. To date, all existing graph classification methods are designed to target one single learning task and require a large number of labeled samples for learning good classification models. In reality, each real-world task may only have a limited number of labeled samples, yet multiple similar learning tasks can provide useful knowledge to benefit all tasks as a whole. In this paper, we formulate a new multi-task graph classification (MTG) problem, where multiple graph classification tasks are jointly regularized to find discriminative subgraphs shared by all tasks for learning. The niche of MTG stems from the fact that with a limited number of training samples, subgraph features selected for one single graph classification task tend to overfit the training data. By using additional tasks as evaluation sets, MTG can jointly regularize multiple tasks to explore high quality subgraph features for graph classification. To achieve this goal, we formulate an objective function which combines multiple graph classification tasks to evaluate the informativeness score of a subgraph feature. An iterative subgraph feature exploration and multi-task learning process is further proposed to incrementally select subgraph features for graph classification. Experiments on real-world multi-task graph classification datasets demonstrate significant performance gain.

Original languageEnglish
Article number7302040
Pages (from-to)715-728
Number of pages14
JournalIEEE Transactions on Knowledge and Data Engineering
Volume28
Issue number3
DOIs
Publication statusPublished - 1 Mar 2016
Externally publishedYes

Keywords

  • Graph Classification
  • Multi-task Learning
  • Regularization
  • Subgraph Features
  • Supervised Learning

Cite this

Pan, Shirui ; Wu, Jia ; Zhu, Xingquan ; Zhang, Chengqi ; Yu, Philip S. / Joint structure feature exploration and regularization for multi-task graph classification. In: IEEE Transactions on Knowledge and Data Engineering. 2016 ; Vol. 28, No. 3. pp. 715-728.
@article{76a7f2cd5b4f4bcdb1ab8b8136769740,
title = "Joint structure feature exploration and regularization for multi-task graph classification",
abstract = "Graph classification aims to learn models to classify structure data. To date, all existing graph classification methods are designed to target one single learning task and require a large number of labeled samples for learning good classification models. In reality, each real-world task may only have a limited number of labeled samples, yet multiple similar learning tasks can provide useful knowledge to benefit all tasks as a whole. In this paper, we formulate a new multi-task graph classification (MTG) problem, where multiple graph classification tasks are jointly regularized to find discriminative subgraphs shared by all tasks for learning. The niche of MTG stems from the fact that with a limited number of training samples, subgraph features selected for one single graph classification task tend to overfit the training data. By using additional tasks as evaluation sets, MTG can jointly regularize multiple tasks to explore high quality subgraph features for graph classification. To achieve this goal, we formulate an objective function which combines multiple graph classification tasks to evaluate the informativeness score of a subgraph feature. An iterative subgraph feature exploration and multi-task learning process is further proposed to incrementally select subgraph features for graph classification. Experiments on real-world multi-task graph classification datasets demonstrate significant performance gain.",
keywords = "Graph Classification, Multi-task Learning, Regularization, Subgraph Features, Supervised Learning",
author = "Shirui Pan and Jia Wu and Xingquan Zhu and Chengqi Zhang and Yu, {Philip S.}",
year = "2016",
month = "3",
day = "1",
doi = "10.1109/TKDE.2015.2492567",
language = "English",
volume = "28",
pages = "715--728",
journal = "IEEE Transactions on Knowledge and Data Engineering",
issn = "1041-4347",
publisher = "IEEE, Institute of Electrical and Electronics Engineers",
number = "3",

}

Joint structure feature exploration and regularization for multi-task graph classification. / Pan, Shirui; Wu, Jia; Zhu, Xingquan; Zhang, Chengqi; Yu, Philip S.

In: IEEE Transactions on Knowledge and Data Engineering, Vol. 28, No. 3, 7302040, 01.03.2016, p. 715-728.

Research output: Contribution to journalArticleResearchpeer-review

TY - JOUR

T1 - Joint structure feature exploration and regularization for multi-task graph classification

AU - Pan, Shirui

AU - Wu, Jia

AU - Zhu, Xingquan

AU - Zhang, Chengqi

AU - Yu, Philip S.

PY - 2016/3/1

Y1 - 2016/3/1

N2 - Graph classification aims to learn models to classify structure data. To date, all existing graph classification methods are designed to target one single learning task and require a large number of labeled samples for learning good classification models. In reality, each real-world task may only have a limited number of labeled samples, yet multiple similar learning tasks can provide useful knowledge to benefit all tasks as a whole. In this paper, we formulate a new multi-task graph classification (MTG) problem, where multiple graph classification tasks are jointly regularized to find discriminative subgraphs shared by all tasks for learning. The niche of MTG stems from the fact that with a limited number of training samples, subgraph features selected for one single graph classification task tend to overfit the training data. By using additional tasks as evaluation sets, MTG can jointly regularize multiple tasks to explore high quality subgraph features for graph classification. To achieve this goal, we formulate an objective function which combines multiple graph classification tasks to evaluate the informativeness score of a subgraph feature. An iterative subgraph feature exploration and multi-task learning process is further proposed to incrementally select subgraph features for graph classification. Experiments on real-world multi-task graph classification datasets demonstrate significant performance gain.

AB - Graph classification aims to learn models to classify structure data. To date, all existing graph classification methods are designed to target one single learning task and require a large number of labeled samples for learning good classification models. In reality, each real-world task may only have a limited number of labeled samples, yet multiple similar learning tasks can provide useful knowledge to benefit all tasks as a whole. In this paper, we formulate a new multi-task graph classification (MTG) problem, where multiple graph classification tasks are jointly regularized to find discriminative subgraphs shared by all tasks for learning. The niche of MTG stems from the fact that with a limited number of training samples, subgraph features selected for one single graph classification task tend to overfit the training data. By using additional tasks as evaluation sets, MTG can jointly regularize multiple tasks to explore high quality subgraph features for graph classification. To achieve this goal, we formulate an objective function which combines multiple graph classification tasks to evaluate the informativeness score of a subgraph feature. An iterative subgraph feature exploration and multi-task learning process is further proposed to incrementally select subgraph features for graph classification. Experiments on real-world multi-task graph classification datasets demonstrate significant performance gain.

KW - Graph Classification

KW - Multi-task Learning

KW - Regularization

KW - Subgraph Features

KW - Supervised Learning

UR - http://www.scopus.com/inward/record.url?scp=84962418082&partnerID=8YFLogxK

U2 - 10.1109/TKDE.2015.2492567

DO - 10.1109/TKDE.2015.2492567

M3 - Article

VL - 28

SP - 715

EP - 728

JO - IEEE Transactions on Knowledge and Data Engineering

JF - IEEE Transactions on Knowledge and Data Engineering

SN - 1041-4347

IS - 3

M1 - 7302040

ER -