forum.alglib.net

Hi all...

I'm using your great libraries (for VisualBasic) mainly the LEVENBERG-MARQUARDT-LIKE METHOD FOR NON-LINEAR OPTIMIZATION.
But how can I make the initial guessing of the parameters ?

Any algorithms or ideas ?

Thanks :)

You can try to guess them from problem properties - this is the best advice known to whole numerical community so far :) Really, answer is problem-specific and there are no general recommendations which will work in any case. If you don't know what to choose, you may try several runs from random points and choose best solution.

Currently, I'm inputting the initial guess manually. But I was trying to find a automated way of guessing the initial parameters, after the X & Y inputs are received from the user.
I'm trying to solve Weibull model : y = a-b*exp(-cx^d)

Everything works fine (till now :)), but the initial guessing needs to be entered manually. But I found this software doing the initial guessing automatically based on the X & Y input values : http://www.curveexpert.net/doc/curveexpert/basic/modeling.html#nonlinear-regression



Thanks... :)

Nothing prevents you from making similar initial guesses - if you know analytic form of function :)