BP神经网络源代码(basic).doc

《BP神经网络源代码(basic).doc》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、PrivateSubCommand5_Click() DimtempNumAsDoubleFactTimes=0Forloopt=0Totimes-1'学习次数     FactTimes=FactTimes+1tempNum=0IfFactTimes<2000Then'学习率的变化前大后小    alpha=0.5    beta=0.5Else    IfFactTimes>2000AndFactTimes<5000Then          alpha=0.4          beta=0.

2、4    Else          alpha=0.3          beta=0.3    EndIfEndIf '开始神经网络计算'/*---Beginningofneuralcomputting---*/Forloopl=1ToLearnExampleNums '学习模式个数     '/*forwardcomputting*/     '/*inputlayer*/     Fori=1ToInputUnitNums          OutofInputLayer(i)=inData

3、s(loopl,i)     Nexti     '/*hidelayer*/     Fori=1ToHideUnitNums          inival=CDbl(0)          Forj=1ToInputUnitNums                inival=inival+w_InputHide(i,j)*OutofInputLayer(j)          Nextj          inival=inival+Cw_Hide(i)          OutofHide

4、Layer(i)=Sigmf(inival)     Nexti     '/*outputlayer*/     Fori=1ToOutUnitNums          inival=0#          Forj=1ToHideUnitNums                inival=inival+w_HideOut(i,j)*OutofHideLayer(j)          Nextj           inival=inival+Cw_Out(i)           Outo

5、fOutLayer(i)=Sigmf(inival)     Nexti     '/*---Backpropagation---*/     '/*deltacaclculate*/     Error=0#     Fori=1ToOutUnitNums                wk=OutofOutLayer(i)                wkb=Teacher(loopl,i)-wk                '计算每个学习模式中各个输出结点的误差平方和           

6、     Error=Error+wkb*wkb                 DEL_Out(i)=wkb*wk*(1#-wk)     Nexti     '/*deltacaclculate*/     Fori=1ToHideUnitNums           inival=0#           Forj=1ToOutUnitNums                 inival=inival+(DEL_Out(j)*w_HideOut(j,i))           Nextj  

7、         wk=OutofHideLayer(i)           DEL_Hide(i)=inival*wk*(1#-wk)     Nexti      '/*updatingforweightsfromHideLayer*/     Fori=1ToOutUnitNums           DCw_Out(i)=alpha*DEL_Out(i)           Forj=1ToHideUnitNums                 Dw_HideOut(i,j)=alpha

8、*DEL_Out(i)*OutofHideLayer(j)               Nextj         Nexti         '/*updatingforweightsfromInputLayer*/         Fori=1ToHideUnitNums               DCw_Hide(i)=beta*DEL_Hide(i)               Forj=1ToInputUnitNums