欢迎来到天天文库
浏览记录
ID:47689794
大小:86.00 KB
页数:18页
时间:2019-10-23
《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
此文档下载收益归作者所有