using lsqnonlin function returns not a vector

I am trying to find 10 parameters value using lsqnonlin. My function returns a 10*10 matrix. each column is the set of data that has to be summed and squared and give one parameter value.Is it possible to do it with lsqnonlin? my function is as follows:

function r = calcResiduals(parVector, settingsStruct)

for i = 1 : size(settingsStruct.expConditions,1) % loop through all experiments

    conc = getconc(settingsStruct.expConditions(i,:),parVector);
  for j=1:10
    r1 = conc(:,j)-settingsStruct.measurementData(:,i);
    r(:,j)=r1;
  end
end

end

i am calling lsqnonlin as follows:

settingStruct.expConditions=[5 5 50];
settingStruct.measurementData=[5;4;3;2;1;0;0.9;0.8;0.7;0.6];
parVector=[1 1 1 1 1 1 1 1 1 1];
lsqnonlin(@calcResiduals,parVector,[],[],[],settingStruct)

 |* _i am gtting error as follows :_ *|

Warning: Failure at t=2.879119e+000.  Unable to meet integration tolerances without
reducing the step size below the smallest value allowed (7.105427e-015) at time t. 
> In ode45 at 309
  In getconc at 21
  In calcResiduals at 5
  In snls at 333
  In lsqncommon at 175
  In lsqnonlin at 237
  In ModelDiscrimination at 68 
Error using horzcat
CAT arguments dimensions are not consistent.

Error in getconc (line 27)
    conc=[cout1(:,1) cout2(:,1) cout3(:,1) cout4(:,1) cout5(:,1) cout6(:,1) cout7(:,1)
    cout8(:,1) cout9(:,1) cout10(:,1)];

Error in calcResiduals (line 5)
    conc = getconc(settingsStruct.expConditions(i,:),parVector);

Error in snls (line 333)
            newfvec = feval(funfcn{3},xcurr,varargin{:});

Error in lsqncommon (line 175)
    [xC,FVAL,LAMBDA,JACOB,EXITFLAG,OUTPUT,msgData]=...

Error in lsqnonlin (line 237)
[xCurrent,Resnorm,FVAL,EXITFLAG,OUTPUT,LAMBDA,JACOB] = ...

Answer 1

[Matt J]

lsqnonlin doesn't care what shape your residual array, r, has. It will always be internally columnized to r(:), see

http://www.mathworks.com/help/optim/ug/matrix-arguments.html

As for the error, it is because the different column vectors cout?(:,i) are not all the same length in

    conc=[cout1(:,1) cout2(:,1) cout3(:,1) cout4(:,1) cout5(:,1) cout6(:,1) cout7(:,1)
    cout8(:,1) cout9(:,1) cout10(:,1)];

So, they cannot be concatenated.

[Devyani]

i dont knw why is tht nt coming... bcoz i tried to solve that getconc function alone then it is working properly.... but whe i combine with the calcResiduals function it is not working... any suggestions???

[Matt J]

... any suggestions???

The debugger

http://www.mathworks.com/help/matlab/matlab_prog/debugging-process-and-features.html#brqxeeu-182

[Devyani]

i changed the tolerance for ode to 1e-2 then it is solving and showing results... thanks

[Matt J]

@PATHAN,

Your question looks to be completely unrelated to this thread, so you should probably start your own. However, a common reason why the initial point is a local min is because of quantization operations like round(), fix(), floor() in your objective function. There are rules about the kind of objective functions you can have.