PyGMO for Winpython

[QCode]

Hi,

Just wondering if this is possible to include the excellent package from European Agency Space :

PYGMO

http://esa.github.io/pygmo/install.html

Optimization tools

Thanks

[stonebig]

hi,

Current choice of optimisation and operation research packages in WinPython is probably not optimal, and I'm wide open to suggestions.

Any suggestion should :

- be python 3.4/3.5 compatible, in wheel package format

- provide some nice links on  examples and tutorials the proposed package

- maybe solves the 99 pences problem in winpython_checker.ipynb (so we can replace Pulp): http://nbviewer.jupyter.org/github/winpython/winpython_afterdoc/blob/master/docs/Winpython_checker.ipynb

On Python3, Pygmo is only provided in 64 bit and in a msi package.

==> It doesn't fit, as it is.

[QCode]

Hi,

Compiling instruction is here:

http://esa.github.io/pygmo/install.html#compiling-and-installing-under-windows

https://github.com/esa/pagmo/issues/169

It has been installed with Enthought Canopy.

Also, quality of the solver is very high.

Kind regards

Thanks for considering it.

[stonebig]

"CMake 2.8.6, Python 2.7, Boost 1.47.0 and MinGW with Msys 1.0.11 (all installed in C:)" 

You'll have to wait for PyGMO to be more Python3/Wheel ready: for sure it will, by 2020.

[Denis Akhiyarov]

pygmo is GPL licensed.

I would recommend dakota, scipy, pyomo for any optimization work.

[stonebig]

I see someone built a version of pygmo for anaconda 2.7 https://anaconda.org/ccordoba12/pygmo/files

[stonebig]

pygmo seems able to solve my diet and pennies problems... have to figure out the re-write of pulp example

On Tuesday, August 16, 2016 at 1:26:16 AM UTC+2, Denis Akhiyarov wrote:

[QCode]

What would be the steps to migrate from Anaconda Windows PYGMO to WinPython ?

Can we just copy/paste the folder ?

[stonebig]

You may indeed try some copy/paste, when you will have identified where PYGMO objects are.

Then maybe a winpython\scripts\make_winpython_movable.bat may be necessary.

The chance of success are slim, but you can try, no kitten will be armed.

[stonebig]

at the moment, moving to cvxopt/cvxpy seems the possible move.

[Denis Akhiyarov]

cvx is great package (i took the class), but the license is again GPL.

Disciplined convex programming is only required at design time of optimization routines.

Then you can switch to whatever nonlinear solver that you like, e.g. ipopt.

[stonebig]

I don't take attention on the licence of each package in WinPython.

Having a package maintained, working well, on Windows, and in WinPython way-of-running is already a very limiting set of criterias,