Function Error not understood

I wrote the following function:

function diff = fit_simp(x,X,Y)

% This function is called by lsqnonlin.
% x is a vector which contains the coefficients of the
% equation.  X and Y are the option data sets that were
% passed to lsqnonlin.

A=x(1);
B=x(2);
n=x(3);

diff = (A+B*(X^n)) - Y;
end

Then the following script;

>> X=xlsread('QS2T.xls',9,'V638:V1835');
>> Y=xlsread('QS2T.xls',9,'W638:W1835');
>> X0=[800 1537 0.1448];
>> options = optimset('Largescale','off');
>> x=lsqnonlin(@fit_simp,X0,[],[],options,X,Y);

I get the error below, don't know what is wrong. All this was given in matlab examples in the website.PLs help, *|??? Error using ==> mpower Matrix must be square.

Error in ==> fit_simp at 12 diff = (A+B*(X^n)) - Y;

Error in ==> lsqnonlin at 202 initVals.F = feval(funfcn{3},xCurrent,varargin{:});

Caused by: Failure in initial user-supplied objective function evaluation. LSQNONLIN cannot continue.|*

Answer 1

[Honglei Chen]

You may want to use X.^n to achieve element-wise operation

[Shalini]

Thanks it works- but still not over!

So the function is;

I added X.^n as you said (please can you explain what it actually does?)

I do not mind running non linear Lavenberg Marquardt or trust region algorithm.

I get this warning and error.Please can you again help?

>> X=xlsread('QS2T.xls',9,'V638:V1835');

Y=xlsread('QS2T.xls',9,'W638:W1835');

X0=[800 1537 0.1448];

>> options = optimset('Largescale','off');

>> x=lsqnonlin(@fit_simp,X0,[],[],options,X,Y);

Warning: Options LevenbergMarquardt and LargeScale will be ignored in

a future release. Running Levenberg-Marquardt method. To run the

Levenberg-Marquardt method without this warning, set option Algorithm

to 'levenberg-marquardt' instead.

> In optim\private\lsqncommon at 60

In lsqnonlin at 240

Maximum number of function evaluations exceeded;

increase options.MaxFunEvals

[Honglei Chen]

It just does element wise operation. Most likely your X is a vector, so you can only raise each element in the vector to a certain power, that's why you need to use .^ operator.

As to the warning from Levenberg Marquardt, the message is preparing for a future change. You may want to set the Algorithm field in your option variable to 'levenberg-marquanrdt'.

[Shalini]

I'm very very new to matlab...Where do I set the algorithm field and how?

That option variable I just coped from a mathworks help webpage.I shall eb grateful if you tell me / write how and what to set?

I also would like to try Trust region

Please help..

[Honglei Chen]

in command line, type

>> option

and you'll see all the fields. To set algorithm field,

>> option.Algorithm = 'levenberg-marquanrdt'

[Shalini]

Thanks and will do that..(I do not have matlab access for the next 2 hours).What abt the error cocerning MAxFunctions Eval?

[Shalini]

I mean the error/warning: Maximum number of function evaluations exceeded;

How to increase this? I tried it by typing in the command line;

increase options.MaxFunEvals=60000 but does not help...??

[Honglei Chen]

this might suggest that the function didn't converge but I'm not sure what's wrong. You may need to examine your data.