Joint Structure Feature Exploration and Regularization for Multi-Task Graph Classification. Pan, S., Wu, J., Zhu, X., Zhang, C., & Yu, P., S. IEEE Transactions on Knowledge and Data Engineering (TKDE), 28(3):715-728, IEEE Computer Society, 3, 2016. doi abstract bibtex 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.
@article{
title = {Joint Structure Feature Exploration and Regularization for Multi-Task Graph Classification},
type = {article},
year = {2016},
keywords = {Graph Classification,Multi-task Learning,Regularization,Subgraph Features,Supervised Learning},
pages = {715-728},
volume = {28},
month = {3},
publisher = {IEEE Computer Society},
day = {1},
id = {f46e588c-385f-3617-84f5-5712f6bbc9b3},
created = {2016-04-29T05:47:43.000Z},
accessed = {2016-04-29},
file_attached = {false},
profile_id = {079852a8-52df-3ac8-a41c-8bebd97d6b2b},
last_modified = {2022-04-10T12:11:14.735Z},
read = {false},
starred = {false},
authored = {true},
confirmed = {true},
hidden = {false},
citation_key = {Pan2016a},
folder_uuids = {2327f56c-ffc0-4246-bac0-b9fa6098ebfb,a5961aba-0e48-4cf8-9566-d73aaba73d2c},
private_publication = {false},
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.},
bibtype = {article},
author = {Pan, Shirui and Wu, Jia and Zhu, Xingquan and Zhang, Chengqi and Yu, Philip S.},
doi = {10.1109/TKDE.2015.2492567},
journal = {IEEE Transactions on Knowledge and Data Engineering (TKDE)},
number = {3}
}
Downloads: 0
{"_id":"yhfvvSt6qKMgtBbXo","bibbaseid":"pan-wu-zhu-zhang-yu-jointstructurefeatureexplorationandregularizationformultitaskgraphclassification-2016","downloads":0,"creationDate":"2016-09-20T01:55:17.104Z","title":"Joint Structure Feature Exploration and Regularization for Multi-Task Graph Classification","author_short":["Pan, S.","Wu, J.","Zhu, X.","Zhang, C.","Yu, P., S."],"year":2016,"bibtype":"article","biburl":"https://bibbase.org/service/mendeley/079852a8-52df-3ac8-a41c-8bebd97d6b2b","bibdata":{"title":"Joint Structure Feature Exploration and Regularization for Multi-Task Graph Classification","type":"article","year":"2016","keywords":"Graph Classification,Multi-task Learning,Regularization,Subgraph Features,Supervised Learning","pages":"715-728","volume":"28","month":"3","publisher":"IEEE Computer Society","day":"1","id":"f46e588c-385f-3617-84f5-5712f6bbc9b3","created":"2016-04-29T05:47:43.000Z","accessed":"2016-04-29","file_attached":false,"profile_id":"079852a8-52df-3ac8-a41c-8bebd97d6b2b","last_modified":"2022-04-10T12:11:14.735Z","read":false,"starred":false,"authored":"true","confirmed":"true","hidden":false,"citation_key":"Pan2016a","folder_uuids":"2327f56c-ffc0-4246-bac0-b9fa6098ebfb,a5961aba-0e48-4cf8-9566-d73aaba73d2c","private_publication":false,"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.","bibtype":"article","author":"Pan, Shirui and Wu, Jia and Zhu, Xingquan and Zhang, Chengqi and Yu, Philip S.","doi":"10.1109/TKDE.2015.2492567","journal":"IEEE Transactions on Knowledge and Data Engineering (TKDE)","number":"3","bibtex":"@article{\n title = {Joint Structure Feature Exploration and Regularization for Multi-Task Graph Classification},\n type = {article},\n year = {2016},\n keywords = {Graph Classification,Multi-task Learning,Regularization,Subgraph Features,Supervised Learning},\n pages = {715-728},\n volume = {28},\n month = {3},\n publisher = {IEEE Computer Society},\n day = {1},\n id = {f46e588c-385f-3617-84f5-5712f6bbc9b3},\n created = {2016-04-29T05:47:43.000Z},\n accessed = {2016-04-29},\n file_attached = {false},\n profile_id = {079852a8-52df-3ac8-a41c-8bebd97d6b2b},\n last_modified = {2022-04-10T12:11:14.735Z},\n read = {false},\n starred = {false},\n authored = {true},\n confirmed = {true},\n hidden = {false},\n citation_key = {Pan2016a},\n folder_uuids = {2327f56c-ffc0-4246-bac0-b9fa6098ebfb,a5961aba-0e48-4cf8-9566-d73aaba73d2c},\n private_publication = {false},\n 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.},\n bibtype = {article},\n author = {Pan, Shirui and Wu, Jia and Zhu, Xingquan and Zhang, Chengqi and Yu, Philip S.},\n doi = {10.1109/TKDE.2015.2492567},\n journal = {IEEE Transactions on Knowledge and Data Engineering (TKDE)},\n number = {3}\n}","author_short":["Pan, S.","Wu, J.","Zhu, X.","Zhang, C.","Yu, P., S."],"biburl":"https://bibbase.org/service/mendeley/079852a8-52df-3ac8-a41c-8bebd97d6b2b","bibbaseid":"pan-wu-zhu-zhang-yu-jointstructurefeatureexplorationandregularizationformultitaskgraphclassification-2016","role":"author","urls":{},"keyword":["Graph Classification","Multi-task Learning","Regularization","Subgraph Features","Supervised Learning"],"metadata":{"authorlinks":{"pan, s":"https://bibbase.org/service/mendeley/079852a8-52df-3ac8-a41c-8bebd97d6b2b"}},"downloads":0},"search_terms":["joint","structure","feature","exploration","regularization","multi","task","graph","classification","pan","wu","zhu","zhang","yu"],"keywords":["graph classification","multi-task learning","regularization","subgraph features","supervised learning"],"authorIDs":["561c75c18d7cb332200004ce","561c77518d7cb332200004db","5de872e7e66c23df0100005d","5e0176b9219bd5df010000c1","5e114756495520de010000a9","5e12c60370e2c4f201000052","5e199a3b204503de0100007f","5e36614a4b25bbf2010000a4","k6fB9cczWrDbvaPzR"],"dataSources":["mKA5vx6kcS6ikoYhW","AoeZNpAr9D2ciGMwa","fcdT59YHNhp9Euu5k","m7B7iLMuqoXuENyof","Byqq56wkTmdCSSibG","gmNB3pprCEczjrwyo","SRK2HijFQemp6YcG3","6aBwAHJEc5NXgW2bw","dJWKgXqQFEYPXFiST","HPBzCWvwA7wkE6Dnk","uEtXodz95HRDCHN22","ya2CyA73rpZseyrZ8","vpu5W6z2tNtLkKjsj","HmWAviNezgcH2jK9X","ukuCjJZTpTcMx84Tz","AcaDrFjGvc6GmT8Yb","2252seNhipfTmjEBQ"]}