Файл: Лабораторная работа 2 Обучение однослойного перцептрона с несколькими выходами.pdf

17
# trainHeavi(netOut[i],weights,[inp1,layer1,layer2],[delta1,delta2],[Ni,
N1,Nout],Beta,Nju,Mju)
# trainLin(netOut[i],weights,[inp1,layer1,layer2],[delta1,delta2],[Ni,N1
,Nout],Beta,Nju,Mju)
# trainReLU(netOut[i],weights,[inp1,layer1,layer2],[delta1,delta2],[Ni,N
1,Nout],Beta,Nju,Mju) trainHT(netOut[i],weights,[inp1,layer1,layer2],[delta1,delta2],[Ni,N1,
Nout],Beta,Nju,Mju)
Corr[i]=isConverged(netOut[i],layer2,eps)
# Corr[i]=isConvergedLin(netOut[i],layer2,eps)
# Corr[i]=isConvergedReLU(netOut[i],layer2,eps)
Corr1=0 for c in range(Nout): if Corr[c]==Nout:
Corr1+=1 epoch+=1 print('Epoch=',epoch,'No errors=',Corr) # Конец цикла обучения print('Число циклов обучения=',epoch) print('No errors=',Corr) print('Test signals:') for i in range(Ntest): print('Signal-',i,':',metki[i],':') showImage(signal2[i],Nix,Niy) vvod(signal2[i],inp1)
# workLog(inp1,w1,layer1,Alpha)
# workLog(layer1,w2,layer2,Alpha)
# workHeavi(inp1,w1,layer1,Alpha)
# workHeavi(layer1,w2,layer2,Alpha)
# workLin(inp1,w1,layer1,Alpha)
# workLin(layer1,w2,layer2,Alpha)
# workReLU(inp1,w1,layer1,Alpha)
# workReLU(layer1,w2,layer2,Alpha)
# logToHT(inp1) workHT(inp1,w1,layer1,Alpha) workHT(layer1,w2,layer2,Alpha) for j in range(Nout): print('Out-',j,'=',layer2[j]) print() input('Continue.') print('Test signals with noise:') for i in range(Ntest): print('Signal-',i,':',metki[i],':') vvod(signal2[i],inp1) noise(inp1,0.125) showImage(inp1,Nix,Niy)

18
# workLog(inp1,w1,layer1,Alpha)
# workLog(layer1,w2,layer2,Alpha)
# workHeavi(inp1,w1,layer1,Alpha)
# workHeavi(layer1,w2,layer2,Alpha)
# workLin(inp1,w1,layer1,Alpha)
# workLin(layer1,w2,layer2,Alpha)
# workReLU(inp1,w1,layer1,Alpha)
# workReLU(layer1,w2,layer2,Alpha)
# logToHT(inp1) workHT(inp1,w1,layer1,Alpha) workHT(layer1,w2,layer2,Alpha) for j in range(Nout): print('Out-',j,'=',layer2[j]) print() input('Press .')