MAT之GA：GA优化BP神经网络的初始权值、阈值，从而增强BP神经网络的鲁棒性

输出结果

实现代码

global p    

global t    

global R  

global S1

global S2  

global S  

S1 = 10;

p = [0.01 0.01 0.00 0.90 0.05 0.00;

    0.00 0.00 0.00 0.40 0.50 0.00;

    0.80 0.00 0.10 0.00 0.00 0.00;

    0.00 0.20 0.10 0.00 0.00 0.10]';

t = [1.00 0.00 0.00 0.00;

    0.00 1.00 0.00 0.00;

    0.00 0.00 1.00 0.00;

    0.00 0.00 0.00 1.00]';

P_test = [0.05 0    0.9  0.12 0.02 0.02;

         0    0    0.9  0.05 0.05 0.05;

         0.01 0.02 0.45 0.22 0.04 0.06;

         0    0    0.4  0.5  0.1  0;

         0    0.1  0    0    0    0]';

net = newff(minmax(p),[S1,4],{'tansig','purelin'},'trainlm');

net.trainParam.show = 10;

net.trainParam.epochs = 2000;

net.trainParam.goal = 1.0e-3;

net.trainParam.lr = 0.1;

[net,tr] = train(net,p,t);

s_bp = sim(net,P_test)  

R = size(p,1);

S2 = size(t,1);

S = R*S1 + S1*S2 + S1 + S2;

aa = ones(S,1)*[-1,1];

popu = 50;  

initPpp = initializega(popu,aa,'gabpEval',[],[1e-6 1]);  

gen = 100;  

[x,endPop,bPop,trace] = ga(aa,'gabpEval',[],initPpp,[1e-6 1 1],'maxGenTerm',gen,...

                          'normGeomSelect',[0.09],['arithXover'],[2],'nonUnifMutation',[2 gen 3]);

figure(1)

plot(trace(:,1),1./trace(:,3),'r-');

title( 'GA优化BP神经网络，绘制均方误差变化曲线—Jason niu')

hold on

plot(trace(:,1),1./trace(:,2),'b-');

xlabel('Generation');

ylabel('Sum-Squared Error');

figure(2)

plot(trace(:,1),trace(:,3),'r-');

title( 'GA优化BP神经网络，绘制适应度函数变化曲线—Jason niu')

hold on

plot(trace(:,1),trace(:,2),'b-');

xlabel('Generation');

ylabel('Fittness');

[W1,B1,W2,B2,val] = gadecod(x);

net.IW{1,1} = W1;

net.LW{2,1} = W2;

net.b{1} = B1;

net.b{2} = B2;

net = train(net,p,t);

s_ga = sim(net,P_test)