How to handle functions with multiple outputs

Dear matlab users,

I have a function with multiple returns, and I want to handle the single output as anonymous function.Let me explain it a bit more,I have afunction multfunc with outputs f1,and f2 as described below

function [f1,f2]=multfunc(x1,x2)
f1=x1*log(x2);
f2=x2^2;
end

I want to find the extremem of f2 and I gave a try as follws:

[~,f2]=@(x1,x2)multfunc(x1,x2)
then in the error is:Only functions can return multiple values

my intention is to return both functions and want to handle f2,How could I do that?

Thanks

Answer 1

[Rik]

You will need a wrapper function (which you can make fancy if you want, but I made it simple).

f=@(x1,x2)wrapper(x1,x2);
function y=wrapper(x1, x2)
[~,y]=multfunc(x1,x2)
end

[Samuel]

Thank you dear Rik,it may be very silly question but I have tried to find the minmum value of f2 using fminserach function starting from intial value of 10,but coudn't get it work.

f=@(x1,x2)wrapper(x1,x2);
function y=wrapper(x1, x2)
[~,y]=multfunc(x1,x2)
x = fminsearch(y,10)
end

[Rik]

You should only use that wrapper function to extract the second output from your function.

You should also note that fminseach only allows a single input, so you have to further tweak either the wrapper, or the anonymous function.

Despite the warning, this does find the smallest value: x(2)=0

f=@(x)wrapper(x(1),x(2));
x = fminsearch(f,[10,10])
function y=wrapper(x1, x2)
[~,y]=multfunc(x1,x2);
end
function [f1,f2]=multfunc(x1,x2)
f1=x1*log(x2);
f2=x2^2;
end

[Samuel]

Thank you dear Rik,actually my intention is to check the values of f1 optimizing f2.

So I can chnage the function into a single output function as follws:

function f2=multfunc(x1,x2)
f1=x1*log(x2);
f2=x2^2;
end
Then I can handle these function using
obj=@(x1,x2)multfunc(x1,x2)

by doing so,in this workspace the value of f1 is suppressed since only f2 called,but I want to find the first function f1 evaluated after optimizing f2.

Thanks

[Rik]

If you leave multfunc intact you can very easily get the value of f1 for a minimized f2:

f=@(x)wrapper(x(1),x(2));
x = fminsearch(f,[10,10])
f1=multfunc(x(1),x(2));

What is your actual goal? You seem to be changing the code I'm suggestion for no obvious reason.

Answer 2

[madhan ravi]

 [~, f2] = multfunc(x1, x2)

[Samuel]

Thanks for your reply,that is how to call and return the second function but my objective is to handle the second function,I want to optimize f2,any help for that matter?

[Walter Roberson]

you will need to use a real function to assign the result of multfunc to variables and then return f2 so it can be optimized.

In MATLAB it is not possible to capture the second or later output of a function within an expression: you need a true assignment statement and that requires a true function not an anonymous function.

[Samuel]

Thank you,can you please make it sense with matlab expressions having the above matlab function ?

function [f1,f2]=multfunc(x1,x2)
f1=x1*log(x2);
f2=x2^2;
end

How to return the two function outputs(f1,f2) and optimize the second function(f2) at the same time,

Thanks

[Walter Roberson]

 function v = Out2(f, varargin)
   [v{1}, v{2}] = f(varargin{:});

Now instead of calling multfunc(x1, x2) instead call Out2(@multfunc, x1, x2)

What you get back will be a cell array with the two outputs. You can then extract the entries from the cell array.

What you cannot do at all easily is use something like ga or fmincon to optimize the second value and then at the end have the optimization routine return the optimal x1 x2 and the optimal value for f2 and simultaneously the f1 corresponding to the optimal f2. The optimization routines are not able to handle output to optimize over along with "extra values".

If that is what you want to do, find the f1 corresponding to the optimal f2, then the easiest way is to use a function similar to the above Out2 that returns only the f2 value, and optimize on that, and then once you know the location of the optimum, call multfunc with that location and look at both outputs. This does involve one "extra" call to multfunc, but it is by far the easiest way to code the situation.