人工智能论文20220604 2019PhD Washington Graph Deep Learning Methods and Applications.pdfVIP

WASHINGTONUNIVERSITYINST.LOUIS

SchoolofEngineeringAppliedScience

DepartmentofComputerScienceandEngineering

DissertationExaminationCommittee:

YixinChen,Chair

MichaelAvidan

SanmayDas

RomanGarnett

BrendanJuba

YinjieTang

GraphDeepLearning:MethodsandApplications

by

MuhanZhang

Adissertationpresentedto

TheGraduateSchool

ofWashingtonUniversityin

partialfulfillmentofthe

requirementsforthedegree

ofDoctorofPhilosophy

December2019

St.Louis,Missouri

©2019,MuhanZhang

TableofContents

ListofFiguresv

ListofTablesviii

Acknowledgmentsx

Abstractxiii

Chapter1:Introduction1

1.1GraphDeepLearning1

1.2ABriefHistoryofGraphNeuralNetworks4

1.3GraphNeuralNetworksBasics6

1.4ACategorizationofGraphNeuralNetworks8

1.4.1GNNsfornode-leveltasks9

1.4.2GNNsforgraph-leveltasks18

1.4.3GNNsforedge-leveltasks19

Chapter2:GraphNeuralNetworksforGraphRepresentationLearning22

2.1GraphNeuralNetworksforGraphClassification23

2.1.1Traditionalgraphclassificationmethods:graphkernels23

2.1.2LimitationsofexistingGNNsfo