How can I find a good initial guess for Levenberg-Marquardt algorithm?

Hello, I would like to know if there are any parameters to set a good initial guess for LMA. I want to apply trilateration with approximate distances and according to different documents LMA is a good approach. My problem is that sometimes I don't get good results and I need to apply the algorithm several times with different initial points. Is there an easy way to know which guess will produce good results?

Thanks in advance!

Answer 1

[Prashant Arora]

Hi Ale,

I don't believe there is a "correct" answer for finding the initial guess for an optimization problem. It depends on how the objective function looks like. The simplest technique which you can use to come up with a better initial guess is doing a random search on the objective function and picking the one resulting in minimum value. You can also refer to MATLAB functions like "GlobalSearch" and "MultiStart" to implement such algorithms more efficiently.

https://www.mathworks.com/help/gads/example-finding-global-or-multiple-local-minima.html