資源描述：
《Reducing the Dimensionality of Data with Neural Networks.pdf》由會員上傳分享，免費在線閱讀，更多相關(guān)內(nèi)容在學(xué)術(shù)論文-天天文庫。
1、www.sciencemag.org/cgi/content/full/313/5786/504/DC1SupportingOnlineMaterialforReducingtheDimensionalityofDatawithNeuralNetworksG.E.Hinton*andR.R.Salakhutdinov*Towhomcorrespondenceshouldbeaddressed.E-mail:hinton@cs.toronto.eduPublished28July2006,Scienc
2、e313,504(2006)DOI:10.1126/science.1127647ThisPDFfileincludes:MaterialsandMethodsFigs.S1toS5MatlabCodeSupportingOnlineMaterialDetailsofthepretraining:TospeedupthepretrainingofeachRBM,wesubdividedalldatasetsintomini-batches,eachcontaining100datavectorsan
3、dupdatedtheweightsaftereachmini-batch.Fordatasetsthatarenotdivisiblebythesizeofaminibatch,theremainingdatavectorswereincludedinthelastminibatch.Foralldatasets,eachhiddenlayerwaspretrainedfor50passesthroughtheentiretrainingset.Theweightswereupdatedafter
4、eachmini-batchusingtheaveragesinEq.1ofthepaperwithalearningrateof
