Sensitivity computing for RBF Neural Network

Hi! I have to implement a sensivity analysis in Matlab for RBFNN. Can somebody help me with that? I am using iris dataset for this.

This is my code for NN learning and error computing:

load iris_dataset; 
goal = 0.001; 
spread = 1; 
MN = 50; 
d = 0.6; 
kl1 = irisInputs(:,1:50);
kl2 = irisInputs(:,51:100); 
kl3 = irisInputs(:,101:end);
learn = [irisInputs(:, 1:30) irisInputs(:, 51:80) irisInputs(:, 101:130)]; 
test = [irisInputs(:, 31:50) irisInputs(:, 81:100) irisInputs(:, 131:150)];
output = [irisTargets(:, 1:30) irisTargets(:, 51:80) irisTargets(:, 101:130)]; 
net = newrb(learn, output, goal, spread, MN); 
outputU = sim(net,learn); 
sumU = 0;
errU = 0; 
for i = 1 : size(learn)
 sumU = sumU + (output(i)-outputU(i))^2; 
 if(output(i) ~= outputU(i))
      errU = errU + 1; 
 end
end
errMsqU = sqrt(sumU);
errIloU = errU/150;

I want to compute output sensitivity to input and weight perturbations. On the internet I have seen only mathematical formulas which are complicated.

Can you give me some Matlab code to do that? Thank's in advance.

I see also these topics, but I don't understand how to code that:

http://www.mathworks.com/matlabcentral/answers/5342-sensitivity-analysis-ann

http://www.mathworks.com/matlabcentral/answers/194647-how-to-compute-sensitivity-analysis-in-neural-network-model