Menu

Dopri/Radau vs. GSL

Help
2016-11-28
2017-01-10
  • Maurizio De Pitta'

    Hi folks,
    I have a very simple question. I am playing around with the integrators, and attempt to reproduce the solutions of my ODEs by GSL ODE solvers... However there is substantial lack of match so far between the solutions that these latter give me with respect to the ones provided by Dopri/Radau integrators.

    Could you point out please what exactly these integrators are in terms of numerical methods. The link on the online tutorial is broken at the moment (see here).

    Thanks.

    M

     
    • Rob Clewley

      Rob Clewley - 2016-11-28

      As the docs point out, both Dopri and Radau originate from Hairer and
      Wanner. While I no longer have access to the wiki source to edit the
      docs online, the FTP of the original code isn't even really necessary.
      The code is all included in PyDSTool, since we compile it. You can
      read more in their two-volume Springer books on integrating ODEs with
      stiff and non-stiff methods. The citations and code are all here:

      http://www.unige.ch/~hairer/software.html

      Depending on the problem, it can be very easy for different
      integrators to diverge in their computed solutions. Sometimes, a lot
      of work has to go into analyzing the situation to choose parameters
      most appropriately to ensure accuracy, especially if over a long
      integration time.

      On Mon, Nov 28, 2016 at 4:37 PM, Maurizio De Pitta'
      mauriziodepitta@users.sf.net wrote:

      Hi folks,
      I have a very simple question. I am playing around with the integrators, and
      attempt to reproduce the solutions of my ODEs by GSL ODE solvers... However
      there is substantial lack of match so far between the solutions that these
      latter give me with respect to the ones provided by Dopri/Radau integrators.

      Could you point out please what exactly these integrators are in terms of
      numerical methods. The link on the online tutorial is broken at the moment
      (see here).

      Thanks.

      M


      Dopri/Radau vs. GSL


      Sent from sourceforge.net because you indicated interest in
      https://sourceforge.net/p/pydstool/discussion/472291/

      To unsubscribe from further messages, please visit
      https://sourceforge.net/auth/subscriptions/

       
  • Maurizio De Pitta'

    Hi Rob,
    thanks for the link!
    I managed by it to recode the Dopri in C/C++ and solve one of my problems (this was few days after your post, but it took a while to reply as I have been relocating to Europe meanwhile...).

    On a follow-up on this issue, could you please point me where to find in the PyDSTool code the routines to compute the Jacobian numerically? I need to use the Radau (after recoding it in C++), but I cannot provide an analytical form for the Jacobian of my system to the integrator (as it would become prohibitive to estimate). PyDSTool allows to integrate by Radau even if the Jacobian is not given, but Hairer's Radau routines require it instead. Accordingly, I believe that somewhere in PyDSTool, there is a routine for numerical estimation of the Jacobian, but I cannot find where this code is.

    Thanks,

    M

     
  • Maurizio De Pitta'

    A follow up on this. Radau's comes already with a code snippet that allows numerical computation of the Jacobian, and this is set by opportune flag upon invocation. Forget my comment above.

     

Log in to post a comment.