Thread: any other method rather than polonomials for curve fitting?

i have y=f(x1,x2)
i could draw 2 curves 1-y wrt x1 and 1- y wrt x2
how i can find the equations?
i know using curve fitting but can i find it in another form rather tha polynomial

http://en.wikipedia.org/wiki/Curve_fitting

The old HP pocket calculators had built in linear, exponential, and log fitting.

I think you can try this on-line fitting tool:

http://www.zunzun.com/

regards,
Alberto

Originally Posted by THUNDERRr

i have y=f(x1,x2)
i could draw 2 curves 1-y wrt x1 and 1- y wrt x2
how i can find the equations?

Hopefully you know the general form of the equation so you only need to find the coefficients.

i know using curve fitting but can i find it in another form rather tha polynomial

Using polynomials is the most general method if you have no clue what the form of the equations is. It takes a lot of processing power to find the coefficients because there are so many if you want to be accurate.

VSWR posted a link. In that link is another link. I usually use the Levenberg-Marquardt algorithm.

Log fitting such as y=x1*ln(x2) or exponential fitting y=x1*exp(x2) assume you now the general relationships or form of the equation.

A polynomial that estimates both x1 and x2 with a 7th order polynomial would require finding 64 coefficients but the computer is doing the work.

Programs such as Scilab(free) and Matlab have optimization routines that make it easy to find the coefficients of a polynomial. There are usually two forms of optimization functions. The L-M algoirthm and the BFGS or L-BFGS algorithm.
Coding functions goes quickly. Waiting for the computer to find the coefficients that minimize the error between the estimating function and the actual data may take a while.

If you do research and understand the physics of the data acquired then you should know the general form the equation. Otherwise use the brute force polynomial approach.