Re: [eigen] Project updated from Eigen2 to Eigen3

On Mon, Apr 8, 2013 at 10:48 AM, Gael Guennebaud <gael.guennebaud@xxxxxxxxx> wrote:

Hi Arnaud,

thank you for the information. I added an entry for BTK in our project
list. IIR filters and writing python wrappers are two quite common
questions on the forum, so I extended the FAQ with your pointers:

http://eigen.tuxfamily.org/index.php?title=FAQ

IIR filters are out of the scope of Eigen, on the other hand as LSQR
solver would be interesting. It's complementary to our CG/BiCGSTAB
solvers. Your code could serve as a basis to write a generic Eigen3
compatible one. A simplex solver would perfectly match too.

cheers,
gael

On Tue, Apr 2, 2013 at 9:15 AM, Arnaud BARRE <arnaud.barre@xxxxxxxxx> wrote:
> Hi there,
>
> I use Eigen since now 5 years in my projects and I would like to thanks to
> you for this great works. It saved me (and still continue) lots of time when
> I needed to do matrix computation to transform data in 3D space.
>
> I updated recently my main Project from Eigen2 to Eigen3 without any
> problem. Maybe it is the time to present it here.
>
> BTK - An open-source Biomechanical ToolKit library, is a BSD library
> implemented in C++ with bindings for Matlab/Octave and Python. A GUI
> software called Mokka is also proposed to visualize/analyze 3D/2D motion
> capture data. The primary goal of this project is to propose a set of tools
> for the analysis of the human body motion which is independent of any
> acquisition system.
>
> With the development of these tools I also implemented some specific
> functions for Eigen.
> - I started a signal processing module for Eigen where some infinite impulse
> response (IIR) filters for 1D signals are already implemented (Butterworth
> filters):
> https://code.google.com/p/b-tk/source/browse/#svn%2FBTK%2Ftrunk%2FUtilities%2Feigen3%2FbtkEigen%2FSignalProcessing
> - For the Python binding i use SWIG and then I developed a SWIG fragment to
> transform an Eigen matrix to a NumPy array:
> https://code.google.com/p/b-tk/source/browse/BTK/trunk/Utilities/SWIG/eigen.i
> - I also implemented an iterative conjugate-gradient-like solver (LSQR:
> still for Eigen2). But I think the solvers included within Eigen are better:
> https://code.google.com/p/b-tk/source/browse/BTK/branches/experimental/Utilities/eigen2/btkEigen/Solver/LSQR.h
> - Finally, It is not yet proposed on the web, but I also implemented a
> downhill simplex algorithm (also named Nelder-Mead and Amoeba) which is
> close to the syntax of the Matlab function 'fminsearch'.
>
> If you think these pieces of code are good enough for the Eigen project, I
> will be very happy to submit some patches. If necessary, I will relicense
> the code to the MPL2 license as I don't see any conflict between the BSD
> license and the MPL2 license (but I'm not a lawyer so if someone disagrees
> let me know).
>
> Arnaud
>
> BTK: https://code.google.com/p/b-tk/
> Mokka : http://b-tk.googlecode.com/svn/web/mokka/index.html
> Mailing-list: https://groups.google.com/forum/?fromgroups#!forum/btk-users
> Twitter: @btkwww
>