Skip to first unread message
Ilhan Polat
Apr 20, 2013, 6:00:25 AM4/20/13
to yal...@googlegroups.com
Dear Johan
As we have communicated a few times, some years ago, I'm a loyal foot soldier of YALMIP for virtually every problem in my research for almost 10 years and let me thank you again for keeping up this monster alive and in good condition. Lately, after so many unfortunate cases of MATLAB bugs/failures/shortcomings/license/price/impossibledebugging..... I'm leaving MATLAB and porting everything to Python. *Die-hard programmers, please, it's just a choice, replace <python> with your favorite language for now*. I've been trying to find a feasible way to bring the convenience of YALMIP to those environments but I can't really see a way yet. I'm trying to find a way to parse complicated LMI descriptions involving quite a number of sdpvars without going down to the cone description or coefficient matrices. I find CVXOPT a very valuable tool in that direction and PICOS comes pretty close though I'm really a beginner for now. I've selfishly picked the semidefinite programming examples but other problems are seemingly supported too.
Now of course, you don't have to share my view or better those of AbandonMATLAB :) so I'd better ask plainly : is there any hope to see YALMIP leaving MATLAB environment (probably not in the near future obviously) to other open-source frameworks? Also in case you see some possibilities for alternative ways, I'd love to hear your view on this issue.
Thanks in advance.
As we have communicated a few times, some years ago, I'm a loyal foot soldier of YALMIP for virtually every problem in my research for almost 10 years and let me thank you again for keeping up this monster alive and in good condition. Lately, after so many unfortunate cases of MATLAB bugs/failures/shortcomings/license/price/impossibledebugging..... I'm leaving MATLAB and porting everything to Python. *Die-hard programmers, please, it's just a choice, replace <python> with your favorite language for now*. I've been trying to find a feasible way to bring the convenience of YALMIP to those environments but I can't really see a way yet. I'm trying to find a way to parse complicated LMI descriptions involving quite a number of sdpvars without going down to the cone description or coefficient matrices. I find CVXOPT a very valuable tool in that direction and PICOS comes pretty close though I'm really a beginner for now. I've selfishly picked the semidefinite programming examples but other problems are seemingly supported too.
Now of course, you don't have to share my view or better those of AbandonMATLAB :) so I'd better ask plainly : is there any hope to see YALMIP leaving MATLAB environment (probably not in the near future obviously) to other open-source frameworks? Also in case you see some possibilities for alternative ways, I'd love to hear your view on this issue.
Thanks in advance.
Johan Löfberg
Apr 20, 2013, 1:33:39 PM4/20/13
to yal...@googlegroups.com
Nope (I would not find it interesting, fun or academically challenging to embark on such a project). I use/develop YALMIP for the stuff I need in my research, and so far MATLAB has been a good-enough platform. I'm too old to learn something new :-)
Ilhan Polat
Apr 20, 2013, 6:37:55 PM4/20/13
to yal...@googlegroups.com
Got it. Thanks ! :)
Erling D. Andersen
May 7, 2013, 7:16:56 AM5/7/13
to yal...@googlegroups.com
MOSEK/Fusion which is a commercial product (free for academics) might be useful It is emphasized at making conic modeling easy in Python. That is of course a more limited scope than Yalmip but anyway. See
Erling
Reply all
Reply to author
Forward
0 new messages