[Sorry for replying to the previous email, but I lost your answer
after reading it.

However I could re-read it at
http://sourceforge.net/mailarchive/forum.php?thread_name=684314.52352.qm%40web110015.mail.gq1.yahoo.com&forum_name=octave-dev
(btw, stange email text...)
]


On 21/09/2009, Luca Favatella <[email protected]> wrote:
[...]
> On 20/09/2009, Adrian Burlacu <[email protected]> wrote:
>> Cool, ok, so it didn't work.
>> Here is how I try to use it in matlab
>> x0=[];

octave-forge linprog is a wrapper aroung glpk.

x0 argument is not supported in octave-forge, and I don't think there
is something analog in glpk (or I can't find it, patches are welcome).
Even in Matlab that argument is used only in particular cases (as far
as I understood).
However, keeping it at [] disables it anyway (I imagine you used it as
a placeholder to input the options argument).


>> options = optimset('TolFun',1e-8);

Same here.
Not supported in octave-forge.
There are some tolerance arguments in glpk, but I think it could be
very difficult to map the tolerance you want to those of glpk (I mean
with the same semantics). Writing tests for this functionality will
also be very difficult (again, time/patches/tests welcome).

So, instead of using
  linprog(f,A,Parab,Aeq,beq,lb,ub,x0,options)
please use
  linprog(f,A,Parab,Aeq,beq,lb,ub)
(or other signatures you read in "help linprog").


[...]
> Could you please write/attach the full octave log (so I can read your
> exact commands and the exact error)?
> (And I remember you to please use svn version of linprog.)

I played with the scenario you emailed (input.mat and output.mat in
the mail in the thread I linked above).


The results are very interesting: your result (x) and the result of
octave-forge linprog are (roughly) the same (168), but:

* octave-forge linprog result is exactly 168, while yours is
1.7053e-13 greater (that is worse); to test it:
 + load input.dat
 + load output.dat
 + (f' * x) + 168  # ans = 1.7053e-13
 + (f' * linprog(f,A,Parab,Aeq,beq,lb,ub)) + 168  # ans = 0

* elements of x are 0.something in yours, all integers numbers using
octave-forge linprog



Please test svn version of linprog and feel free to report your feedback.



Thanks,
Luca Favatella

------------------------------------------------------------------------------
Come build with us! The BlackBerry&reg; Developer Conference in SF, CA
is the only developer event you need to attend this year. Jumpstart your
developing skills, take BlackBerry mobile applications to market and stay 
ahead of the curve. Join us from November 9&#45;12, 2009. Register now&#33;
http://p.sf.net/sfu/devconf
_______________________________________________
Octave-dev mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/octave-dev

Reply via email to