[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[Help-gsl] Possible multiroot shortcoming

From: Remi Magnin
Subject: [Help-gsl] Possible multiroot shortcoming
Date: Wed, 12 Nov 2014 17:40:13 +0100


I am new here so I am sorry I do not know exactly how it works to ask
questions on this mailing list.

I encounter a problem using the multiroot facility, I tried hard to find
solutions anywhere but have not found it yet.
I try to briefly describe my problem. I have a quite complex non linear
system of 6 equations with 6 parameters to determine. For the moment I will
not copy them here because it is a big amount of cosinus and sinus
functions (if you require it I can try to do it later). I use
multiroot_fdf_solver. I have already checked my Jacobian and my functions
multiple times, at this point I do not see any error. My problem is the
-through iteration, if my initial guess is close to the real values, the
parameters(the ones I try to guess) will not update at all, and if I try to
display the dx it has the value "NaN". However the status of
is "success". After about 10 iterations, it stops (still in "success"
status, and it seems it considers it has converged), and give me as a
return the initial guess(but they are not the values I expect).

-the parameters will only be updated if my initial parameters are far from
their theoretical values. Obviously in this case it does not converge to
the good values.

At this point the only problem that looks like what I have is described in
an (old) thread of this ML called "Possible multiroot shortcoming", I give
the link:

My question would be: has anyone already had this kind of problem, and do
we have some news about the quoted thread because it is really similar to
my problem, and I would like to know whether there is a possible "bug" in
the multiroot facility because my function is too complicated, or it is me
that made an error in my code (as I said I checked it over and over, and I
do not find an error yet).

Thank you in advance for your answer, and do not hesitate to ask me
aditionnal informations if needed, your help would be greatly appreciated!

reply via email to

[Prev in Thread] Current Thread [Next in Thread]