Curve Fitting with Individual trial function

Hello,

matlab offers the possibility to approximate numerical data by a function like polynomials, exponential function etc! Is there a way to do this with an individual trial function like f(x)= a*sqrt(x)*(b*x^2+exp(c*x) or something like that. If the number of (x,y) pairs is the same like the number of coefficients to be determined in my trial equotation, it could be done with a nonlinear equotation system but i want the curve it to be approximated!

Thanks for your help

[Star Strider]

See ‘lsqcurvefit’ or ‘nlinfit’. If you don't have access to these, you can use ‘fminsearch’ to do nonlinear curve fitting.

Answer 1

[Tom Lane]

If you are talking about the Curve Fitting Toolbox, try this:

f = @(a,b,c,x) a*sqrt(x).*(b*x.^2+exp(c*x))
y = f(1,2,-3,x) + randn(size(x))/100;
plot(x,f(1,2,-3,x),'b.')
fit(x,y,f)

You could also enter the expression into the cftool window. Notice that I vectorized your expression, for example changing "^" to ".^" to get element-by-element operations on the vector x.

[Christian]

Thank you very much!