MathGroup Archive 2005

[Date Index] [Thread Index] [Author Index]

Search the Archive

Re: Quite Upset with NIntegrate

  • To: mathgroup at smc.vnet.net
  • Subject: [mg54178] Re: Quite Upset with NIntegrate
  • From: "Alan" <info at optioncity.REMOVETHIS.net>
  • Date: Sat, 12 Feb 2005 01:58:25 -0500 (EST)
  • References: <cuhsk4$9i8$1@smc.vnet.net>
  • Sender: owner-wri-mathgroup at wolfram.com

> Should I accept that Mathematica can not handle such a
> numerical integration or is there a way to make Nintegrate working
> better?

Often NIntegrate gets bogged down in 2+dimensional
integrals or even 1D integrals with complicated integrands.
When that happens to me, I usually work my way up to
the computation by first creating an interpolating function for
the integrand and integrating that. That is quite controlled in the sense 
that
most of the time will be spent in sampling your function on
some grid of your choosing and you know exactly how many
points are going to be sampled.  The resulting integration is usually
fast, but not necessarily very accurate. But, at least you then
know how long it takes to sample your integrand at N points.

Then, try the original problem again with AccuracyGoal->2 and
count how many points are sampled and how long it takes.
You should at that point have a good esimate of the
running time/accuracy trade-off.

Hope it helps,
alan

>
> Ismail
> 



  • Prev by Date: Re: VertexStyle determined by List elements
  • Next by Date: upgraded to Fedora Core 3
  • Previous by thread: Re: Quite Upset with NIntegrate
  • Next by thread: Re: Quite Upset with NIntegrate