call function in fmicon optimization failing

Hi guys, I have 4 functions where each one is derived from previous one. When I call the last one (revenue.m) I got an error using fmincon. Any clue? I want to optimize in respect to x0 that is present in each function. Many Thanks! Here is my code:

%% Total out flow equation T = zeros([N,M]); S=zeros([N,M]); for i = 1:M

    R(:,i)=revenue(price(:,1),HH(:,i),ro,g,eff,x0(:,i),N,k1);
    options = optimset('MaxFunEvals',Inf,'MaxIter',50,...
    'Algorithm','interior-point','Display','iter');    
tic 
    [x(:,i),fval(:,i)] = fmincon(@revenue,x0(1:2*N,i),A,b(:,i),Aeq,beq(:,i),LB(:,i),UB(:,i),[],options); %good   
toc
end

with revenue fun being:

        function R=revenue(price,HH,ro,g,eff,x0,N,k1)

            R= -sum(price.*((HH*ro*g*eff).*x0(1:N))/k1);

        end
        % saved as revenue.m

Error msg:

    Not enough input arguments.
Error in revenue (line 5)
            R= -sum(price.*((HH*ro*g*eff).*x0(1:N))/k1);
Error in fmincon (line 535)
      initVals.f = feval(funfcn{3},X,varargin{:});
Caused by:
    Failure in initial objective function evaluation. FMINCON cannot continue.
  end

[Torsten]

fmincon expects "revenue" to be of the form

function R=revenue(x)

where x is the vector of unknowns.

Your function has the form

function R=revenue(price,HH,ro,g,eff,x0,N,k1)

[sensation]

Thanks Torsten. How I am supposed to write then revenue fun if it is calculated using (price,HH,ro,g,eff,x0,N,k1)? Any tip? Many thanks really!

[Torsten]

What is the "x" in your list that fmincon expects ? Is it x0 ?

[sensation]

yes. with initial guess :

inFlow = rand(10,3);

x0 = [inFlow; zeros(size(inFlow))];

Thanks!

Answer 1

[Stephen23]

The fmincon documentation clearly states that the first input " fun is a function that accepts a vector or array x and returns a real scalar f, the objective function evaluated at x". So it only accepts one input, which can be a vector or matrix. How many inputs does revenue have?

You need to parameterize the function, e.g. using an anonymous function and assuming that all of HH, ro, etc. are defined in that workspace, and the first input price is that one that you want to optimize:

fun = @(x) revenue(x,HH,ro,g,eff,x0,N,k1);
fmincon(fun,...)

https://www.mathworks.com/help/optim/ug/passing-extra-parameters.html

https://www.mathworks.com/help/matlab/math/parameterizing-functions.html

[sensation]

Hi Stephen and thanks for your msg. I used eventually as you suggested the wrapper funtion:

 tic 
  [x(:,i),fval(:,i)] = fmincon(@(x0)revenue(price(:,i),HH(:,i),ro,g,eff,x0(:,i),N,k1),x0(1:2*N,i),A,b(:,i),Aeq,beq(:,i),LB(:,i),UB(:,i),[],options);  
 toc

 and it runs for the first time step when M=1, then it crashes because: Index exceeds matrix dimensions. Thanks for any tip!

[Torsten]

... = fmincon(@(x0)revenue(price(:,i),HH(:,i),ro,g,eff,x0,N,k1),...

[sensation]

Thanks. Can you just explain me why we did not have to use every column of x0(:,i) in wrapped fun and only x0? Because we expect x0? Cheers!

[Torsten]

The solution inside fmincon is passed as a one-dimensional vector.

Writing

... = fmincon(@(x0)revenue(price(:,i),HH(:,i),ro,g,eff,x0(:,i),N,k1),...

assumes that x0 inside fmincon is a matrix from which you want to pass the i'th column to "revenue". But this i'th column of a matrix does not exist unless i=1.

Best wishes

Torsten.

[sensation]

Thanks!!!