[R-sig-dyn-mod] lsode vs ode15s; ok with radau

[Thomas Petzoldt]

First of all,

A "Happy New Year" to all useRs and dynamic modellers!

On 1/2/2012 1:45 PM, Faelens Ruben wrote:
>> [...]Unfortunately, there is still a significant computational
> difference:
>>
>> ode15s: 35s
>> radau: 1333s
>> [...]
>
> Please see
> http://www.bioconductor.org/help/course-materials/2010/BioC2010/Efficien
> tRProgramming.pdf for hints on efficient R programming. My model sped up
> by a factor 5 by removing names from matrices and carefully looking at
> which objects were duplicated.

I agree completely. One part of the solution is to find the most 
suitable solver and the other to implement the model in an efficient 
way, see recent threads about:

"R vs. fortran for simulation" and "Vectorizing"

in the archives:

https://stat.ethz.ch/pipermail/r-sig-dynamic-models/2011q4/thread.html


Thomas P.



-- 
Dr. Thomas Petzoldt
Technische Universitaet Dresden
Faculty of Forest, Geo and Hydro Sciences
Institute of Hydrobiology
01062 Dresden, Germany

E-Mail: thomas.petzoldt at tu-dresden.de
http://tu-dresden.de/Members/thomas.petzoldt