資源描述：
《simulation of a perceptron using eloom使用eloom感知器的仿真》由會員上傳分享，免費在線閱讀，更多相關內(nèi)容在應用文檔-天天文庫。
1、SimulationofaperceptronusingeLoomInthetutorial,wewillconstructasingleperceptronnetworkwithtwoinputsusingeLoom.Figure1.Aperceptronneuron.1.Createthenetworkspecificationfile.eLoomsupportsfivetypesofcomponents:Module,Node,Nodearray,Connection,andConnectionarray.For
2、thesimpleperceptronnetwork,therearethreenodesandtwoconnections.Wecangroupthetwoinputnodesasanarray,theconnectionsbetweeninputnodesandoutputnodeasconnectionarray.Amodulewillholdallthesecomponents.Belowshowstheskeletonofthenetworkspecificationfile(refereLoomuser’s
3、manualforhelpofcommands):ms-mnPerceptronn-nnYna-nanX-d1-nc02cnaf-canW-nafrX-ntoYmeEOFAddsimulationfunctionsAlleLoomcomponentshaveuserspecifiedsimulationfunctionsanddatastorage.Fortheperceptron,thereisathresholdfunctionappliedtotheoutputnode,weightfunctionsofthec
4、onnections,andlearningruleofthenetwork.Soweaddfunctionsandthecorrespondingfunctionparameterstothespecification:ms-mnPerceptron-fnPerceptronFunc-nfp5-fp0X-fp1Y-fp20.01-fp3training.dat-fp40n-nnY-fnThresholdNodeFunc-nfp2-fp00.0-fp10.0na-nanX-d1-nc02cnaf-canW-cnfnGr
5、idIndicesAppend-nafrX-ntoY-fnWeightBasicFunc-nfp2-fp00.0-fp11.0meEOFTheperceptronfunctionwillreaddatafromexternalfiles,andadapttheconnectionweightsaccordingsomelearningrule(deltarule).Ithasfiveperameters:-fp0X:theinputnodearray-fp1Y:outputnode-fp20.01:learningra
6、te-fp3training.dat:datafile-fp40:runningmode(learningortestingstage)Theseparametersarechangeable,dependingontheactualimplementationofthefunction.Thresholdnodefunctionisstraightforward.Ittakestwoparameters:thethresholdvalue,andthebias.Theweightfunctionwilladdthew
7、eightedsumofinputstotheactivationoftheoutputnodes.Theweightsofthetwoconnectionsintheperceptronwillberandomlyinitiated.Thetwoparametersdefinetheboundoftherandomlyinitialweights.GridIndicesAppendisanamingfunction,whichnametheconnectionsintheconnectionarrayaccordin
8、gtosomenumericalorder.AdddatastorageWehavetoprovidethedatastorageforthesimulationfunctions.ms-mnPerceptron-fnPerceptronFunc-nfp5-fp0X-fp1Y-fp20.01-fp3training.dat-fp4
