0405图论深度学习应用.pptxVIP

下载本文档

0
0
约5.34千字
约 13页
2026-02-26 发布于山东
举报

0405图论深度学习应用.pptx

OptimizationonGraphsPartI:OutlookandApplications

BackgroundandMotivationGraphConstructionandOptimizationPCBPlacementHalf-centuryresearchesonclassicalgraphmatchingproblem50yearsofresearchonasinglegraphtheoryproblemTSPGCPMFPMCPRobertBixbyProf.atRiceJeromeChaillouxAsst.Prof.atIPParisErlingAndersenAsst.Prof.atSDUMoregraphtheoryproblemchallengesinreal-worldscenarios50yearsofresearchonasinglegraphtheoryproblem

Discrete/UnconstrainedDiscrete/ConstrainedContinuous/UnconstrainedContinuous/Constrained(Deep)NeuralNetworkLogisticRegressionPCAGraphTheoryandCOCO:combinatorialopt.ClusteringGraphCutPathPlanningMatching…BackgroundPullDecisionTreesGaussianMixedModelSupportVectorMachineNa?veBayesProbabilisticTopicModelsContinuousDiscreteUnconstrainedConstrained

Matrix/SequenceGraphStructurePerceptualProblemsGraphTheoryProblemsContinuousUnconstrainedDiscreteConstrainedBackgroundProblemSettingDataFormGeneralNNCOModelsMethodologyInnovation

BackgroundofResearchCVPR’22BestPaper:LearningtoSolveHardMinimalProblemsEJOR’21ASurveybyProf.Bengio:MachineLearningforCONSFmakes$20millioninvestmentinOptimization-focusedAIResearchInstituteOpenScenariosVelocityVolumeVarietyVeracityClassicalComfortZoneRelativeStabilityModerateSizeClassicalProblemMachineLearningClassicalSolversLargeSizeRapidChangeMultipleFormsMoreUncertaintyExactCertainty

GraphMatchingBasedModelFusionICML22InputModelAlignmentModelfusionOutputModelDeepNeuralNetworkFusionviaGraphMatchingwithApplicationstoModelEnsembleandFederatedLearning,ICML22Codeavailableat“/Thinklab-SJTU/GAMF”GitHubrepoQRcode

ModelEnsemblePrediction-basedModelEnsemble:NeedtomaintainallindividualmodelsFusion-basedModelEnsemble:NeedtomaintainonlyonemodelFederatedLearningFLPipeline：EfficientlyaggregatelocalmodelsbyModelFusionLi,Q.,He,B.,andSong,

您可能关注的文档

文档评论（0）

1亿VIP精品文档

更多 >

0405图论深度学习应用.pptxVIP