Graph Classification Using Evolutionary Computation]

《Graph Classification Using Evolutionary Computation]》由会员上传分享，免费在线阅读，更多相关内容在学术论文-天天文库。

1、GAIA:GraphClassificationUsingEvolutionaryComputationNingJinCalvinYoungWeiWangUniversityofNorthCarolinaUniversityofNorthCarolinaUniversityofNorthCarolinaatChapelHillatChapelHillatChapelHillChapelHill,NC,USAChapelHill,NC,USAChapelHill,NC,USAnjin@cs.unc.eduyoungc@c

2、s.unc.eduweiwang@cs.unc.eduABSTRACTtobeabletopredictwhichchemicalcompoundsaretoxicandwhichcomponentsarecharacteristicsofchemicaltoxicityDiscriminativesubgraphsarewidelyusedtodefinethefeature[Helma,2004];biologistsareinterestedinstudyingwhichspaceforgraphclassifi

3、cationinlargegraphdatabases.Severalproteinsareabletobindcertainligandsandwhichcanbeusedtoscalableapproacheshavebeenproposedtominediscriminativetreatdiseases[Bandyopadhyay,2006];computerscientistsseektosubgraphs.However,theirintensivecomputationneedspreventfindou

4、thowtolocatebugsinprogramsbyidentifyingthemfrommininglargedatabases.Weproposeanefficientdiscriminativesubgraphsinprogramflowgraphs[Cheng,2009].methodGAIAforminingdiscriminativesubgraphsforgraphPerformingtheseclassificationtasksbyhandisintractableclassificationin

5、largedatabases.Ourmethodemploysanovelcomputationally,thusincreasingattentionhasbeendevotedinsubgraphencodingapproachtosupportanarbitrarysubgraphdevelopinggraphclassificationmethodsinrecentyears.patternexplorationorderandexploresthesubgraphpatternspaceinaprocessr

6、esemblingbiologicalevolution.Inthismanner,1.1RelatedWorkGAIAisabletofinddiscriminativesubgraphpatternsmuchfasterExistingresearchoftenassumesabinarygraphclassificationtaskthanotheralgorithms.Additionally,wetakeadvantageofparallelwhereatargetgraphsetandabackground

7、graphsetaregivenandcomputingtofurtherimprovethequalityofresultingpatterns.Intheobjectiveistoconstructaclassificationmodelfortheend,weemploysequentialcoveragetogenerateassociationdistinguishingthem.Onestraightforwardsolution[Deshpande,rulesasgraphclassifiersusing

8、patternsminedbyGAIA.2005;Bandyopadhyay,2006]tographclassificationisfirstExtensiveexperimentshavebeenperformedtoanalyzethefindingfrequentsubgraphpatterns[Inokuchi,2000