using genetic algorithm for optimization

i have this error message

??? Subscripted assignment dimension mismatch.

Error in ==> fcnvectorizer at 14
            y(i,:) = feval(fun,(pop(i,:)));

Error in ==> makeState at 47
        Score =
        fcnvectorizer(state.Population(initScoreProvided+1:end,:),FitnessFcn,1,options.SerialUserFcn);

Error in ==> gaunc at 41
state = makeState(GenomeLength,FitnessFcn,Iterate,output.problemtype,options);

Error in ==> ga at 279
        [x,fval,exitFlag,output,population,scores] = gaunc(FitnessFcn,nvars, ...

Error in ==> main at 11
[x,fval,exitflag,output,final_pop]=ga(FitnessFcn,nvar,options);

Caused by:
    Failure in user-supplied fitness function evaluation. GA cannot continue.

Answer 1

[Alan Weiss]

Apparently there is an error in your fitness function. If it isn't too long, please post it so we can see what might be the problem.

And please include your gaoptimset calls and your function call to ga (e.g., x = ga(@main,4,A,b) )

Alan Weiss

MATLAB mathematical toolbox documentation

Answer 2

[Walter Roberson]

Your fitness function is returning your input values x_pop as its first return value. It should not be doing that. The documentation indicates,

fitnessfcn

Handle to the fitness function. The fitness function should accept a row vector of length nvars and return a scalar value.

When the 'Vectorized' option is 'on', fitnessfcn should accept a pop-by-nvars matrix, where pop is the current population size. In this case fitnessfcn should return a vector the same length as pop containing the fitness function values. fitnessfcn should not assume any particular size for pop, since ga can pass a single member of a population even in a vectorized calculation.

So your fitnessfcn should only be returning a single variable, not two variables, and (unless you have Vectorized turned on) return a scalar such as fx_val .

[mado]

do you mean that i should write it function [ fx_val]=PID_objfun_ITAE(x_pop,options)

Answer 3

[mado]

this is the fitness function code

function [x_pop, fx_val]=PID_objfun_ITAE(x_pop,options)

KP=x_pop(1);
KI=x_pop(2);

   simopt = simset('solver','ode45','SrcWorkspace','Current','DstWorkspace','Current');  % Initialize sim options
   [tout,xout,yout] = sim('pid_INC_GA',[0 .1],simopt);
  t =0:0.1:30;

      for i=1:301
  error(i) = (abs(26.3-yout(i)))*t(i);
  error=error*error';
  ISE=sum(error);
  end

           sys_overshoot=max(yout)-26.3; % compute the overshoot

        alpha=10;beta=1;
        fx_val=ISE*beta+sys_overshoot*alpha;

[mado]

and this is the main

%initialize genetic

      nvar=2;
FitnessFcn = @PID_objfun_ITAE

    options=gaoptimset('populationsize',30,'Generations',100,....
    'SelectionFcn', @selectionroulette,'EliteCount',6)

%options=gaoptimset('Vectorize','on'); %fasten time %options = gaoptimset('InitialPop', final_pop); %start algorithm %______________________________________

    [x,fval,exitflag,output,final_pop]=ga(FitnessFcn,nvar,options);
fprintf('Genetic Algorithm values : Kd = %7.5f  Kp = %7.5f  Ki = %7.4f ',x(1),x(2),x(3));

[Alan Weiss]

Please format your answers so that we can more easily read them. Use the {}Code button.

Alan Weiss

MATLAB mathematical toolbox documentation

[mado]

can't do more ,excuse me

[Walter Roberson]

Please put a breakpoint in your PID_objfun_ITAE function at the last line, and check size(ISE) and size(sys_overshoot) . If sys_overshoot turns out not to be a scalar, please check size(yout)

[mado]

size(yout) =1000 and error still there what do you mean by breakpoint

[Walter Roberson]

http://blogs.mathworks.com/loren/2010/03/19/debugging-approaches/

Note: every iteration of your "for" loop, you are overwriting error and ISE. After your instruction error=error*error'; then "error" is being replaced by a scalar. It is unlikely that that is what you want. Perhaps you want that instruction and the ISE assignment to be after the "for i" loop.

[mado]

ok i know i put it outside for loop but where is the error here?

[Walter Roberson]

Put a breakpoint at the fx_val assignment. Run the program. When it stops, show us

size(ISE)
size(sys_overshoot)
size(ISE*beta+sys_overshoot*alpha)

If those values look reasonable, use the command

dbcont

to continue. See if you again stop at the same line, or if you get an error first. If you again stop at the same line then the problem is not triggered immediately so you should test those sizes and continue executing until you see something odd.

My speculation is that at some point an empty matrix is being generated.

[mado]

size of them all are [1 1] and i typed dbcont it gives the error message directly

Answer 4

[mado]

i make the function return one variable fx-val but the program is iterating as it is in a loop and restart its work without stopping i don't think its the for loop of error cause i removed it and it gives error message this is a part of it "??? Error using ==> PID_objfun_ITAE at 28 Error due to multiple causes.

Error in ==> validate>@(x)fitness(x,FitnessFcnArgs{:}) at 136 fitness = @(x) fitness(x,FitnessFcnArgs{:});