fitting vector valued functions

I see Matlab has many different ways to fit a scalar valued function (for example

z = f(x,y)

but is there anything that can help me fit a vector valued function

[u, v] = f(x,y)

Note that the coefficients to the fit would be multiplied by a vector, in the equation of f. for example, a term could be a*[x y]

This causes an error in the scalar fitting routines because the coefficient 'a' to be found by fitting can only be multiplied by a scalar.

Answer 1

[Shashank Prasanna]

From your questions I appears that you want to perform Multivariate Regression where the response is m dimensional

Y = F(X) where Y belongs to |Rm and x belongs to |Rn You can do this using MVREGRESS:

http://www.mathworks.com/help/stats/mvregress.html#btkr6ds-3

[Shashank Prasanna]

Of course this is still linear regression.

Answer 2

[Matt J]

LSQCURVEFIT can do multi-variate non-linear fitting.

http://www.mathworks.com/help/optim/ug/lsqcurvefit.html

You need the Optimization Toolbox, of course.

Answer 3

[Matt J]

There are also FEX tools like this, which don't require any toolboxes.

http://www.mathworks.com/matlabcentral/fileexchange/10093-fminspleas

It works best when you have only a few nonlinear parameters, though.