Re: [Help-glpk] Indicator constraints

[Nigel Galloway]

The following:

http://hutchies.iconbar.com/bigm.html

reaches an interesting conclusion:

Some would say that the Big M method is utterly needless and fiddly and
begs to be destroyed from human knowledge. Others would counter that it
removes the need for two objective functions in two-stage simplex and
that its premises are actually rather elegant. I am inclined to agree
with the former party; however, since it is a proven fact that most
mathematicians are insane intelligence-masochists, forever seeking
techniques that will cause their minds yet more pain, it seems unlikely
that this method will disappear in the foreseeable future. I hope this
explanation of its workings has given the reader the reserves of courage
and understanding needed for them to weather the difficult period until
the happy day when Big M is once again no more than a letter of the
alphabet.

I think (I won't comment on agreement or otherwise with the above, other
than to say 1) former sounds good to me; 2) I assume he means the Arabic
Alphabet rather than the Bulgarian, aka azbuka):

GLPK already implements part of a surreal number system, with minimum
mip gap, which allows one to define a number smaller than any positive
number in the number system but greater than zero. Whats required is a
number greater than any number in the number system, such that any
operation performed on it results in a number outside the real number
system.

-- 
  Nigel Galloway
  address@hidden


On Fri, Dec 16, 2011, at 09:29 PM, Xypron wrote:
> Hello Andrew,
> 
> users of GLPK have had often had problems with the accuracy of big M
> formulations.
> 
> In CPLEX big M formulations can be replaced by indicator constraints.
> This is a constraint switched on and off by a binary.
> http://yalma.fime.uanl.mx/cplex11-manual/Content/Optimization/Documentation/CPLEX/_pubskel/XPlatform/User_man747.html
> http://www.ampl.com/MEETINGS/TALKS/2011_11_Charlotte_TD18.pdf
> 
> I guess this would also be a good feature for GLPK.
> 
> Best regards
> 
> Xypron
> 
> _______________________________________________
> Help-glpk mailing list
> address@hidden
> https://lists.gnu.org/mailman/listinfo/help-glpk
> 

-- 
http://www.fastmail.fm - Choose from over 50 domains or use your own