alphabet1.m

人工神经网络资料

clear
[alphabet,targets] =prprob1;
S1 = 10;
[R,Q] = size(alphabet);
[S2,Q] = size(targets);
P = alphabet;
net = newff(minmax(P),[S1  S2],{'logsig' 'logsig'},'traingdx','learngd');
for i=1:36
    subplot(6,6,i)
    plotchar(alphabet(:,i));
end
pause
close all;
%  Training Without Noise
P = alphabet;
T = targets;
net.performFcn = 'mse';
net.trainParam.goal = 0.01;
net.trainParam.lr=0.001;
net.trainParam.show = 100;
net.trainParam.min_grad=1.0e-008;
net.trainParam.epochs = 10000;
net.trainParam.mc = 0.90;
[net,tr] = train(net,P,T);



%  Training With Noise
netn = net;
netn.trainParam.goal = 0.005;
netn.trainParam.epochs = 1000;
T1 = [targets targets targets targets];
for pass = 1:10
P1 = [alphabet, alphabet, ...
      (alphabet + randn(R,Q)*0.1), ...
      (alphabet + randn(R,Q)*0.2)];
[netn,tr] = train(netn,P1,T1);
end

%  Training Without Noise
P2= alphabet;
T2= targets;
net.performFcn = 'mse';
net.trainParam.goal = 0.01;
net.trainParam.show = 100;
net.trainParam.epochs = 5000;
net.trainParam.min_grad=1.0e-008;
net.trainParam.mc = 0.95;
[net,tr] = train(net,P2,T2);

j=round(rand*26+1);
noisyAZ = alphabet(:,j)+randn(35,1) * 0.2;
figure(1)
plotchar(noisyAZ);
A2 = sim(net,noisyAZ);
A2 = compet(A2);
answer = find(compet(A2) == 1);
figure(2)
plotchar(alphabet(:,answer));