the fmincon sets a wrong dimension for my decision variables and that gets me errors on dimension disagreement.

I have a nonlinear objective function with nonlinear constraints that I would like to minimize. I pass parameters and functions for the calculation of the objective function and constraints. So basically my fmincon looks like this:

[x,FVAL]=fmincon(@(x)myOF(x,PSR,PEBU,PWE,PLE,PAE,PHW,PHT,PMT,PST,PHR,PIVTT,PIVTQ,PIVTH,PIVTE,PIVTWE,PIVTLE,PIVTAE,PIVTHW,PCT,PICTE,PMSCp,POR,PER,PSEp,PICTWE,PICTLE,PICTAE,PICTHW,AHW,DHW,VHW,AWE,DWE,VWE,AAE,DAE,VAE,ALE,DLE,VLE,UP,I,J,K,T),x0,[],[],[],[],[],[],@(x)constraints(x,PSR,PEBU,PWE,PLE,PAE,PHW,PHT,PMT,PST,PIVTH,PIVTE,PIVTWE,PIVTLE,PIVTAE,PIVTHW,PCT,PICTE,PICTWE,PICTLE,PICTAE,PICTHW,IB,D,I,J,K,T),options);

'myOF' and 'constraints' are both functions of x and many parameters and the all the parameters are imported from excel. all the parameters are matrices of four dimensions.

I define my x0 to be a zero matrix with four dimensions but after I run the optimization I get error on dimension match because the x matrix is automatically set to be a 2 dimensional. when I run 'myOF' and 'constraints' manually I get no errors.. I am wondering how I can control the dimension of the final x matrix so that Matlab can continue the calculations without having problems with the non-matching dimensions.

Answer 1

[Alan Weiss]

fmincon takes the dimensionality of the problem from x0. Give the correct x0 and you will have no dimensionality problems.

Also, I suggest that you save yourself some writing and possible copy-paste errors by writing

 fun = @(x)myOF(x,PSR,...); % The ... stand for the rest of the parameters
 nonlcon = @(x)constraints(x,PSR,...)

and then call fmincon as

 [x,FVAL] = fmincon(fun,x0,[],[],[],[],[],[],nonlcon,options)

Alan Weiss

MATLAB mathematical toolbox documentation

[saman Tayari]

Thanks a lot for the quick response. I defined x0 as a four dimensional(5x3x4x8) matrix but after I run the model it turns into a two dimensional(5x96).

[Alan Weiss]

I am surprised that you are having this problem. Perhaps you should contact technical support, in case this is a bug in Optimization Toolbox.

As a workaround, you can define x0 as a column vector

 x0 = x0(:);

and then, inside your objective and nonlinear constraint functions, reshape x to the size you want

 x = reshape(x,5,3,4,8);

Then at the end of your function files reshape it back to a column vector

 x = x(:);

Alan Weiss

MATLAB mathematical toolbox documentation

[saman Tayari]

This reshaping seems to be working. Though I need to make sure everything is in good shape after that doing that trick . Plus I will contact the technical support as the 5x96 dimension seems strange. I REALLY APPRECIATE YOUR HELP I felt so desperate for few days.