深度学习大数据分析外文文献翻译

《深度学习大数据分析外文文献翻译》由会员上传分享，免费在线阅读，更多相关内容在工程资料-天天文库。

1、外文文献翻译原文及译文标题：PrototypingaGPGPUNeuralNetworkforDeep・LearningBigDataAnalysis作者:AlcidesFonseca,BrunoCabral期干H：EigDataResearch,卷8：50・56页年份:2017原文PrototypingaGPGPUNeuralNetworkforDeep-LearningBigDataAnalysisAlcidesFonseca,BrunoCabralAbstractBigDataconcernswithlarge-volumecomplexgrowingdata・Give

2、nthefastdevelopmentofdatastorageandnetwork,organizationsarecollectinglargeever-growingdatasetsthatcanhaveusefulinformation.Inordertoextractinformationfromthesedatasetswithinusefultime,itisimportanttousedistributedandparallelalgorithms.Onecommonusageofbigdataismachinelearning,inwhichcollecte

3、ddataisusedtopredictfuturebehavior.Deep・LeamingusingArtificialNeuralNetworksisoneofthepopularmethodsforextractinginformationfromcomplexdatasets.Deep-learningiscapableofmorecreatingcomplexmodelsthantraditionalprobabilisticmachinelearningtechniques・Thisworkpresentsastep-by-stepguideonhowtopro

4、totypeaDeep・LeamingapplicationthatexecutesbothonGPUandCPUclusters.PythonandRedisarethecoresupportingtoolsofthisguide.ThistutorialwillallowthereadertounderstandthebasicsofbuildingadistributedhighperformanceGPUapplicationinafewhours・Sincewedonotdependonanydeep-learningapplicationorframework—w

5、euselow-levelbuildingblocks一thistutorialcanbeadjustedforanyotherparallelalgorithmthereadermightwanttoprototypeonBigData.Finally,wewilldiscusshowtomovefromaprototypetoafullyblownproductionapplication.Keywords:Big-data;Deep-learning;Prototyping;GPGPU;Cluster;ParallelprogrammingIntroductionDee

6、pLearningreferstotheusageofArtificialNeuralNetworks(ANNorNN)withseveralhiddenlayersusedfordatawithahighdimensionality.AcommonexampleandbenchmarkforDeepLearningisimageclassificationfromtheImageNetdataset.ANNscanbeusedforclassificationtasks,withseveralapplicationsinindustry,businessandscience

7、・Examplesofapplicationsincludecharacterrecognitioninscanneddocuments,predictingbankruptcyorhealthcomplications.AutonomousdrivingalsomakesheavyuseofANNs.AnANNbeginswithrandomweights,practicallydecidingeverythingatrandom・BytrainingtheANNwithseveralexisting