how to write function file containing 2 objectives as the 2 seperated output functions of a neural network with 2 outputs.

i seperated 2 outputs of the single function {i.e, for "f = @(t) sim(net,t')"} by using

row1 = @(v) v(1,:);
first_output_function =@(t) row1(sim(net,t'));

similarly i got

row2 =@(v) v(2,:);
second_output_function =@(t) row2(sim(net,t'));

Now i want to create a function file with them as 2 objectives {i.e, 1st output function is 1st objective and 2nd output is 2nd objective) and access it as a objective function in genetic algorithm

note: i know that we can actually write

objFcn = @(t) sim(net,t')

though we have multiple outputs, but still i need this so that sometimes i should be able to change one of the objectives.

I hope i was able to explain my question clearly.

Answer 1

[Walter Roberson]

Go ahead and do that. Just be warned that if you call upon both objective functions then sim() would be run twice, once for the first objective function and once for the second objective function.

You may wish to look at http://blogs.mathworks.com/pick/2015/03/27/cache-your-nuts-to-eat-later/

csim = @(t) cachedcall(@sim, net, t');
row1 = @(v) v(1,:);
row2 = @(v) v(2,:);
first_output_function =@(t) row1(csim(t));
second_output_function = @(t) row2(csim(t));

[sethu]

i created a function file like this

function f = myfunc(t);
csim = @(t) cachedcall(@sim,net,t');
r1 =@(v) v(1,:);
r2 =@(v) v(2,:);
f(1) = @(t) r1(csim(t));
f(2) =@(t) r2(csim(t));
end

Then i gave in command

>> objFcn = @myfunc;
>> nvars= 3;
>> LB = [40,10,6];
>> UB = [86,135,36];
>> [X,FVAL] = gamultiobj(objFcn,nvars,[],[],[],[],LB,UB)

Then i am getting an error mentioning that "non-scalar arrays of function handles are not allowed , use cell arrays instead." Can you please tell me where have i done wrong?

[Walter Roberson]

gamultiobj does not take vector of objective functions as an argument. gamultiobj takes as an argument a single function that returns a vector of objective values.

It is recommended that you write your objective function to be vectorized; see http://www.mathworks.com/help/gads/computing-objective-functions.html . In that case, the output should have one column per objective function.

Basically you should be invoking something like

[X,FVAL] = gamultiobj(@(t) sim(net,t').', nvars, [], [], [], [], LB, UB)

[sethu]

thanks for the reply, suppose my objFcn=@(t) sim(net,t') has two outputs(1st row and 2nd row) then gamultiobj tries to minimize both of them. if i want to maximize both the outputs , then i can use "objFcn =@(t) 1./sim(net,t');" but what can i do if i want to minimize one of the output and maximize the other one. i hope i was able to express my doubt.