Hi Mauro, Thank you for your reply!
Sure, we can discuss some ideas of how to fit a Taylor ODE integrator within ODE.jl and sketch a proposal, how do you think would be the best way to discuss this? I'm eager to learn Julia as much as I can and implement some other methods of integration for ODE.jl if you want me to, I have worked before with implicit, symplectic Runge-Kutta methods also (8th and 12th order). Out of looking for a numerically precise, robust, ODE solver, which could integrate some celestial mechanics problems, is how we came across the Taylor method. I haven't participated in GSoC also, but I have been reading a bit about the process, and I think the next step is to submit a proposal on March 14th-25th. Also, I talked today to Luis Benet (chair of my advisory committee) about the GSoC proposal to contribute to ODE.jl and told me that if it gets approved he could get some resources so that I could go to work with you guys for a couple of weeks! Lastly, who do you think would be also interested to participate as a mentor for this project? I'll be happy to work with whoever you guys decide is the best mentor for me! Best, Jorge On Saturday, February 27, 2016 at 3:51:26 PM UTC-6, Mauro wrote: > > Hi Jorge, > > thanks for your interest! I think this would be a welcome addition to > ODE.jl and I would like to see how this somewhat non-standard method > fits within the framework of ODE.jl. I think this could make a strong > application. > > A couple of things: I never participated in GSOC, so I don't know how to > proceed. Whilst I did update the GSOC ODE.jl proposal, there may also > be other suitable mentors. I guess that is something to be discussed. > > Cheers, > Mauro > > @Ilya: Jorge is a PhD student of the authors of TaylorSeries.jl. > > On Fri, 2016-02-26 at 09:21, pere...@gmail.com <javascript:> wrote: > > Hi Mauro, > > > > I would like to submit a proposal to work on the ODE.jl package, > > for the GSoC. From my undergraduate and master thesis I have > > experience with the Taylor method for solving ODEs (ie., based on Taylor > > series expansions). This is a variable order, variable step > > size method, which uses automatic differentiation > > techniques in order to reach high order integration methods (30th, 40th > > order) > > which enable machine-epsilon precision with very competitive speeds. > > I think the Taylor method is important to include in the ODE.jl package, > > as it is very versatile and precise. > > > > Besides the utility of the Taylor method for ODEs integration, a DAEs > > solver can > > also be implemented using the Taylor models framework. > > > > I would be very happy to contribute to the ODE.jl package! > > > > Best regards, > > > > On Thursday, February 11, 2016 at 7:56:45 AM UTC-6, Mauro wrote: > >> > >> > >> It is desirable to have ode-solvers which are pure Julia. Both to cut > >> down on dependencies and to allow easy hacking and development. > >> Further, Sundials.jl will not work with generic Julia datatypes (e.g. I > >> think Julia sparse matrices are not supported for Jacobians). Thus, > >> ODE.jl is to stay and to be improved on. > >> > >> The currently ongoing work of which I'm aware is: > >> https://github.com/JuliaLang/ODE.jl/pull/49 > >> https://github.com/JuliaLang/ODE.jl/pull/72 > >> > >> Needed work is: > >> - more solvers > >> - a unified code structure/API > >> - parallelism(?) > >> > >> I'll try and update the GSoC description. > >> >
