       Re: Limits on solving large nonlinear system

• To: mathgroup at smc.vnet.net
• Subject: [mg80861] Re: Limits on solving large nonlinear system
• From: Joerg Schaber <schaber at molgen.mpg.de>
• Date: Tue, 4 Sep 2007 03:49:54 -0400 (EDT)
• References: <fbdmq2\$pbo\$1@smc.twtelecom.net>

For large systems of nonlinear equations you might wnat to use rather
global algorityhms rather than local ones, like the ones implemented in
FindRoot.
FindRoot is problematic because in complex nonlienar landscapes the
solution heavily depends on the inital values. If you don't know where
your parameters are roughly located you're basically lost.
Try NMinimize with SimulatedAnnealing or DifferentialEvolution, for
instance. You will not get the best solution but you will get one, even
if it takes some times.

hope that helps,

joerg

> Hi,
>
>
>
> I am using Mathematica to solve nonlinear systems of equations,
>
> without success, because I run out of memory.
>
> So I try to find out where the limits are for FindRoot:
>
> size of the system, degree of the equations and so on.
>
> I actually did not find any information about this.
>
>
>
> In all my equations the degree of one variable is always one,
>
> but I have a lot of expressions where variables are multiplied
>
> like
>
>    x1*x2*x3*x4 = 0
>
>    x1*x2*x3+x1*x4 = 0.
>
>
>
> Does anyone know more about the limits?
>
>
>
> Regards,
>
> Tomislav
>
>
>

• Prev by Date: Re: ===\$MaxMachineNumber}
• Next by Date: Re: ndsolve ndsv problem
• Previous by thread: Limits on solving large nonlinear system
• Next by thread: Re: rotating rings illusion