Algebraic Problem in writing the function for LSQNONLIN tool in the Optimization Toolbox

Hello,

I am using lsqnonlin for fitting a non linear equation to the model

z=Ax+(B*log(y))+(C/y)+(D*x*log(1+y))

X matrix is 10x2 double matrix. 1st column contains values for x and 2nd column contains values for y.

x(1) and x(2) represent x and y respectively in the model.

The function definition to be used in lsqnonlin is as follows:

function F=myfun(x,X,Z) k=1:10; F=(x(1)*X(k,1))+(x(2)*log(X(k,2)))+(x(3)/X(k,2))+(x(4)*X(k,1)*log(1+X(k,2)))-Z(k); end

I have all the values for the X and the Z matrix.

The problem is that whenever I am trying to call the myfun function, the error which it shows is:

??? Error using ==> plus Matrix dimensions must agree.

Error in ==> myfun at 3 F=( x(1) * X(k,1) )+( x(2) * log(X(k,2) )+( x(3) / X(k,2) )+( x(4) * X(k,1) * log(1+X(k,2) ) ) - Z(k);

Please clear this up.

Thank you.

[Andrei Bobrov]

may be try use 'lsqcurvefit'

Answer 1

[Andrei Bobrov]

use function lsqcurvefit

eg:

A = .5;
B = 4.1;
C = .6;
D = .6;
x = (.01:.01:1)';y = (.01:.01:1)';
z = A*x+(B*log(y))+(C./y)+(D*x.*log(1+y))+.1*rand(size(x));

%solution:

f = @(a,X) a(1) * X(:,1) + a(2) * log(X(:,2)) +  a(3) ./ X(:,2) + a(4) * X(:,1) .* log(1+X(:,2) );

abcd = lsqcurvefit(f,[0 0 0 0],[x y],z)

Answer 2

[Sargondjani]

the message simply tells you that if youo want to add two matrices they have to be of the same size, and apparently that is not the case

(split the equation in parts if you want to find out where it goes wrong)

it might be that your Z is a row vector and not a column vector.

i also see a future error message: if you want to devide a number by a vector you have to use './' instead of '/' (after the x(3))

[Rohan Khadatkar]

But isn't it that I am extracting the value from the matrix and then using it in the equation? Why should it give an error?

Moreover, Z matrix is a column matrix.

[Sargondjani]

but you also add x(1)*X(k,1) + x(2)*log(X(k,1))...

that's adding two matrices (vectors). andrei's correction might do the trick though...

[Andrei Bobrov]

give code calling error