Re: Why keeping clicking in the same method yields different results?

Diego R Apr 13, 2010

Okay, so, while any sort of auto RMS minimizator algorithm is not implemented, one can still look for the smallest value manually after a few clicks, right?



--- In find_orb@yahoogroups.com, Bill J Gray <pluto@...> wrote:
>
> Hi Diego,
>
> > I wonder why should I click several times in any given method
> > (e.g. Full Step or even Auto) to obtain smaller RMS?
> >
> > Some cases, you even get oscillating values.
>
> True. That usually happens when the orbit is poorly defined,
> though sometimes in other cases.
>
> Find_Orb has a lot of code that attempts to find where the
> "real" minimum RMS is. Certain cases still confuse it; in some,
> it doesn't even converge to the oscillating values you mention.
>
> Best thing to do then is to switch to use of Herget steps.
> These are somewhat more stable. Even then, all is not perfect;
> you can get cases where things diverge badly.
>
> Lately, I've been working a bit with a "downhill simplex"
> method. (This is currently implemented only in the console version
> of Find_Orb, which is where I frequently test things that eventually
> make their way into the Windows GUI version of Find_Orb.) The math
> behind this method is described at
>
> http://www.fizyka.umk.pl/nrbook/c10-4.pdf
>
> This appears to get good results in almost every case, even
> when started with a completely stupid idea about what the orbit
> might be. It's a little computationally intensive, though.
>
> Probably a better method would be statistical ranging, but
> I've been promising that one for a while now and just haven't
> had the time to do it right.
>
> -- Bill
>