[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [Swarm-Support] 3d diffusion in abm
From: |
Scott Christley |
Subject: |
Re: [Swarm-Support] 3d diffusion in abm |
Date: |
Mon, 20 Feb 2012 21:01:19 -0600 |
Hi Rick,
It can depend upon whether the number of particles diffusing is small and thus
stochastic effects are important, or whether the number of particles is large
enough so you can use a continuous representation. In the former case there
isn't a whole lot that can be done, but there are advanced approximation
algorithms like tau-leaping to gain speedup. In the latter case then a
diffusion PDE can be written and numerical methods used to solve the PDE over
the grid. It is still expensive for 3D, but a coarser grid could possibly be
used depending upon the fidelity needed, plus many numerical methods (e.g.
Runge-Kutta) can be parallelized. For example in one of my models [1] I use
GPU to solve 3D diffusion and get 100x speedup over CPU code.
cheers
Scott
[1] http://www.biomedcentral.com/1752-0509/4/107
On Feb 20, 2012, at 5:04 PM, Rick Riolo wrote:
> hi,
> some colleagues have a question about doing (3d) diffusion in an abm; they
> say:
>
> we have an ABM with lots of diffusion going on and now it is in 3-D!!
> so, it runs VERY VERY SLOWLY...we know that we need to do
> some kind of adaptive refinement, but that is a project in itself...
> do you happen to know of someone who has had to deal with this
> issue or who may know about how to implement that in ABM world?
>
> Anyone have or know of ABMs with lots of diffusion, and thus
> use some special techniques to speed things up?
>
> thanks for any pointers/refs/etc.
> -r
>
> Rick Riolo address@hidden
> Center for the Study of Complex Systems (CSCS)
> 321a West Hall
> University of Michigan Ann Arbor MI 48109-1107
> Phone: 734 763 3323 Fax: 734 763 9267
> http://cscs.umich.edu/~rlr
>
> _______________________________________________
> Support mailing list
> address@hidden
> http://www.swarm.org/mailman/listinfo/support