[SciPy-dev] GSoC weekly report

dmitrey openopt@ukr....
Sat Jun 30 07:31:42 CDT 2007


Hi all,
see
http://openopt.blogspot.com/
for my reports.

Briefly:

1. QP class created (the only one connected qp solver is cvxopt_qp, I 
wrote cvxopt_mosek qp binding but i didn't tested the one because of 
problems with mosek binary libs)

see
from scikits.openopt import QP
help(QP)
for more details.

2. General constrained NLP solver "lincher" (LINearisation solver from 
CHERkassy town) had been written. However, it requires QP solver, and 
cvxopt QP is GPL-licensed. In future I guess I can add native OO QP 
solver (+bonus - handling of QC- Quadratic Constraints, but 
positive-defined only, elseware the problem could be NP-hard), it will 
be able of handling problems with nVars up to 1000, including 
ill-conditioned ones and other difficulties. However, it requires 
consulting with some people from my department, and they are very busy 
for now with their own work.
Also, lincher requires line-search solver, currently I use fminbound 
from scipy, but I intend to remove the dependence later.

see
from scikits.openopt import NLP
help(NLP)
for more details.

or see the notes at http://openopt.blogspot.com/

Regards, D.



More information about the Scipy-dev mailing list