[R-sig-dyn-mod] New R packages supporting compiled code

[Daniel Kaschek]

Dear Karline,

I had a look at your ccSolve package and it looks very nice. I would 
like to comment on two things:

1. Jacobians: When implementing my first BVPs with compiled code I 
found it especially tedious to compute and implement the jacobian of 
the system. Therefore, I implemented a function jacobianSymb() in my 
cOde package that symbolically computes the jacobian for a function 
defined as character vector. If you had plans to implement something 
similar, you might just have a look at my code and use it if you like.

2. Parameter estimation in ODEs: Our group has been working on this 
topic for many years now. In our experience, optimization with 
derivatives computed from finite differences of the model solutions is 
quite unstable and time consuming. Optimization is improved a lot when 
you compute the sensitivities along with the model states. In 
http://journals.plos.org/plosone/article?id=10.1371/journal.pone.0074335 
you find a comparison of different optimization stategies 
(deterministic vs. stochastic, finite differences vs. sensitivities), 
Figures 5, 6 and 7.

Best regards,
Daniel




---------------------------------
Daniel Kaschek
Institute of Physics
Freiburg University

Room:  210
Phone: +49 761 2038531


On Sa, Apr 18, 2015 at 10:52 , Karline Soetaert 
<Karline.Soetaert at nioz.nl> wrote:
> Well, it appears that many people had the same idea:
> See
> 
> https://github.com/karlines/ccSolve/
> 
> which does something similar for Fortran, and C code - and also 
> extends some base R functions to support compiled code.
> 
> Karline
> 
> -----Original Message-----
> From: R-sig-dynamic-models 
> [mailto:r-sig-dynamic-models-bounces at r-project.org] On Behalf Of Tim 
> Keitt
> Sent: vrijdag 17 april 2015 23:45
> To: r-sig-dynamic-models at r-project.org
> Subject: [R-sig-dyn-mod] New R package
> 
> I've recently been using Rcpp + ODEINT for some projects where I 
> needed a lot of speed and flexibility in model specification 
> (variable dimensions, mixing discrete and continuous time functions). 
> It gave me the idea to write a small R package that allows one to 
> specify the ODE system in a C++ snippet, which is combined with 
> boilerplate code and compiled using Rcpp.
> 
> Have a look: https://github.com/thk686/odeintr
> 
> THK
> 
> --
> http://www.keittlab.org/
> 
> 	[[alternative HTML version deleted]]
> 
> _______________________________________________
> R-sig-dynamic-models mailing list
> R-sig-dynamic-models at r-project.org
> https://stat.ethz.ch/mailman/listinfo/r-sig-dynamic-models
> 
> _______________________________________________
> R-sig-dynamic-models mailing list
> R-sig-dynamic-models at r-project.org
> https://stat.ethz.ch/mailman/listinfo/r-sig-dynamic-models