欢迎来到天天文库
浏览记录
ID:34575875
大小:1.08 MB
页数:130页
时间:2019-03-08
《learning deep architectures for AI (2009)》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、RFoundationsandTrendsinMachineLearningVol.2,No.1(2009)1–127c2009Y.BengioDOI:10.1561/2200000006LearningDeepArchitecturesforAIByYoshuaBengioContents1Introduction21.1HowdoWeTrainDeepArchitectures?51.2IntermediateRepresentations:SharingFeaturesandAbstractionsAcrossTasks71.3DesiderataforLearn
2、ingAI101.4OutlineofthePaper112TheoreticalAdvantagesofDeepArchitectures132.1ComputationalComplexity162.2InformalArguments183LocalvsNon-LocalGeneralization213.1TheLimitsofMatchingLocalTemplates213.2LearningDistributedRepresentations274NeuralNetworksforDeepArchitectures304.1Multi-LayerNeuralN
3、etworks304.2TheChallengeofTrainingDeepNeuralNetworks314.3UnsupervisedLearningforDeepArchitectures394.4DeepGenerativeArchitectures404.5ConvolutionalNeuralNetworks434.6Auto-Encoders455Energy-BasedModelsandBoltzmannMachines485.1Energy-BasedModelsandProductsofExperts485.2BoltzmannMachines535.3
4、RestrictedBoltzmannMachines555.4ContrastiveDivergence596GreedyLayer-WiseTrainingofDeepArchitectures686.1Layer-WiseTrainingofDeepBeliefNetworks686.2TrainingStackedAuto-Encoders716.3Semi-SupervisedandPartiallySupervisedTraining727VariantsofRBMsandAuto-Encoders747.1SparseRepresentationsinAuto
5、-EncodersandRBMs747.2DenoisingAuto-Encoders807.3LateralConnections827.4ConditionalRBMsandTemporalRBMs837.5FactoredRBMs857.6GeneralizingRBMsandContrastiveDivergence868StochasticVariationalBoundsforJointOptimizationofDBNLayers898.1UnfoldingRBMsintoInfiniteDirectedBeliefNetworks908.2Variationa
6、lJustificationofGreedyLayer-wiseTraining928.3JointUnsupervisedTrainingofAlltheLayers959LookingForward999.1GlobalOptimizationStrategies999.2WhyUnsupervisedLearningisImportant1059.3OpenQuestions10610Conclusion110Acknowledgments112References113RFoundationsandTrendsinMachineLearningVol.2,No.1(
7、2009)1–127c2009Y.BengioDOI:10.1561/2200000006LearningDeepArchitecturesforAIYoshuaBengioDept.IRO,Universit´edeMontr´eal,C.P.6128,Montreal,Qc,H3C3J7,Canada,yoshua.bengio@umontreal.caAbstractTheoreticalresultssuggestthatinordertolearnthekindofcom-plicatedfunctio
此文档下载收益归作者所有