Mein MATLAB Forum

net = network;
net.numInputs    = 1;
net.numLayers    = 2;


% input layer
net.inputs{1}.size = 2;

% hidden layer
net.layers{1}.size = 25;
net.layers{1}.transferFcn   = 'logsig';
net.layers{1}.initFcn       = 'initwb';
net.biasConnect(1)          = 1;

% output layer
net.layers{2}.size = 2;     % 2 outputs
net.layers{2}.transferFcn   = 'logsig';
net.layers{2}.initFcn       = 'initwb';
net.biasConnect(2)          = 1;
% Connect layers
net.inputConnect(1)     = 1;
net.layerConnect(2,1)   = 1;
net.outputConnect(2)    = 1;

% init weights/bias
 net.inputWeights{1,1}.initFcn       = 'rands';
 net.inputWeights{1,1}.learnFcn      = 'learngdm';
 net.layerWeights{2,1}.initFcn       = 'rands';
 net.layerWeights{2,1}.learnFcn      = 'learngdm';
 net.biases{1}.initFcn       = 'rands';
 net.biases{1}.learn         = 1;
 net.biases{1}.learnFcn      = 'learngdm';
 net.biases{2}.initFcn       = 'rands';
 net.biases{2}.learn         = 1;
 net.biases{2}.learnFcn      = 'learngdm';


net.initFcn             = 'initlay';
net.inputWeights{1,1}.learnFcn = 'learngdm';
net.inputWeights{1,1}.learnParam.lr = 0.5;

net.performFcn  = 'mse';
% net.trainFcn    = 'trainlm';
% net.trainParam.goal=1e-4;

net.adaptFcn = 'adaptwb';

net = init(net);

% view(net)

% init
P0 = {[input(1,1); input(1,2)] };
T0 = {[target(1,1);target(1,2)]};
 
[net,a,e,pf] = adapt(net,P0,T0);
 
for i = 1:size(dDyn,1)-3

        P = {[input(i,1); input(i,2)] };
        T = {[target(i,1);target(i,2)]};
   
        e = 100; n = 0;
        while(mse(e)>0.1 && n<200) % Abbruchbedingung e = 0.1
            [net,a,e,pf] = adapt(net,P,T, pf);  
            % net.adaptParam.passes = 100;
            mse(e)
            n = n+1;
        end
    end  
end