IMPLEMENTATION OF ARTIFICIAL NEURAL NETWORK ON FPGA

IMPLEMENTATION OF ARTIFICIAL NEURAL NETWORK ON FPGA

ID:37658105

大小:318.45 KB

页数:11页

时间:2019-05-27

IMPLEMENTATION OF ARTIFICIAL NEURAL NETWORK ON FPGA_第1页
IMPLEMENTATION OF ARTIFICIAL NEURAL NETWORK ON FPGA_第2页
IMPLEMENTATION OF ARTIFICIAL NEURAL NETWORK ON FPGA_第3页
IMPLEMENTATION OF ARTIFICIAL NEURAL NETWORK ON FPGA_第4页
IMPLEMENTATION OF ARTIFICIAL NEURAL NETWORK ON FPGA_第5页
资源描述:

《IMPLEMENTATION OF ARTIFICIAL NEURAL NETWORK ON FPGA》由会员上传分享,免费在线阅读,更多相关内容在行业资料-天天文库

1、IMPLEMENTATIONOFARTIFICIALNEURALNETWORKONFPGA12Dr.RezaRaeisi,ArminKabir1IndianaStateUniversity,Indiana;Email:reza@indstate.edu2IndianaStateUniversity,Indiana;Email:akabir1@indstate.edu1.IntroductionAnArtificialNeuralNetwork(ANN)isaninformationprocessingparadigm,whichisinspiredbytheway

2、biologicalnervoussystems,suchasthebrain,processinformation.Thekeyelementofthisparadigmisthenovelstructureoftheinformationprocessingsystem.Scientiststrytoimitatethebrain’scapabilitieswithANN(Haykin,1999).FPGAbasedreconfigurablecomputingarchitecturesarewellsuitedtoimplementANNbecausethe

3、parallelstructureofFPGAmatchesthetopologiesofANN.TheabilitytoeasilyreconfigureFPGAmakesthedesignlessexpensivethanpre-designedhardware,anditalsogivestheuserstheopportunitytochangetheweightsbasedontheirneeds(Ienne,1995).ImplementedANNonFPGAcanbeusedtoworkwithdifferentintegratedcircuits.

4、Forexample,ANNisusedincellphonestodecreasethenoise,inthisapplicationANNshouldbeimplementedonthehardwarebecauseofthesmallsizeofcellphonesandalsohighspeedthatisneededforthisapplication.AmericanSocietyforEngineeringEducationMarch31-April1,2006–IndianaUniversityPurdueUniversityFortWayne(I

5、PFW)2006Illinois-IndianaandNorthCentralJointSectionConference2.ArtificialNeuralNetworksSomeofthemostimportantpropertiesofthehumanbrainthatscientistsliketoimitatearetrial-and-errorandpatternlearning.WithANN,theyhavecomeastepclosertoreachingthisgoal.ANNisusedtosolvemanydailyproblemsindi

6、fferentareasofscience.SomeofthemanyapplicationsofANNareimageprocessing,signalprocessing,systemcontrol,anddatamining.TheabilityofANNtolearnbyexamplemakesitveryflexibleandpowerfulfornumerousapplications.NeuronsarethemainpartofeachANN.Eachneuroncanhavecountlessinputs;however,itwillonlyha

7、veoneoutputvalue,asillustratedinFigure2.1.X1..W1....F(?XiWi).....WnXnFigure2.1NeuronStructureTheoutputofeachneuroniscalculatedbythefollowingformula:Output=F(?Xi*Wi)Formula2.1WhereXiistheinputtotheneuron,andWiistheweightofeachinput.TheseweightsAmericanSocietyforEngineeringEducationMarc

8、h31-A

当前文档最多预览五页,下载文档查看全文

此文档下载收益归作者所有

当前文档最多预览五页,下载文档查看全文
温馨提示:
1. 部分包含数学公式或PPT动画的文件,查看预览时可能会显示错乱或异常,文件下载后无此问题,请放心下载。
2. 本文档由用户上传,版权归属用户,天天文库负责整理代发布。如果您对本文档版权有争议请及时联系客服。
3. 下载前请仔细阅读文档内容,确认文档内容符合您的需求后进行下载,若出现内容与标题不符可向本站投诉处理。
4. 下载文档时可能由于网络波动等原因无法下载或下载错误,付费完成后未能成功下载的用户请联系客服处理。