[Help-glpk] Re: RE: Floating point precision?

[glpk xypron]

Hello Laura,

> The d[i]'s are percentages, so technically they can range between 0-1.
Please, replace
subject to Bin_d1 {i in I1}: d[i] <= 100*BIN2[i];
subject to Bin_d2 {i in I1}: d[i] >= k - 100*(1-BIN2[i]);
by
subject to Bin_d1 {i in I1}: d[i] <= BIN2[i];
subject to Bin_d2 {i in I1}: d[i] >= k * BIN2[i];

This will result in better scaling and could give you up to two
extra digits of precision.

> A few of the d's are around 10^(-8), so those I can write off as rounding
> errors. In a few instances, the "errors" are higher (0.0005), and that
> should not happen.

Some tolerance in GLPK are as high as 1e-5, e.g.:

double tol int (default: 1e-5)
Absolute tolerance used to check if optimal solution to the current LP
relaxation is integer feasible.

Best regards

Xypron

-------- Original-Nachricht --------
> Datum: Thu, 22 Jul 2010 15:55:40 -0400
> Betreff: RE: Floating point precision?

> Thanks for the reply, I didn't catch that detail.
> The d[i]'s are percentages, so technically they can range between 0-1.
> Taking the '100' out however didn't fix the issue.
> 
> A few of the d's are around 10^(-8), so those I can write off as rounding
> errors. In a few instances, the "errors" are higher (0.0005), and that
> should not happen.
> 
> Any other ideas?
> 
> -Laura
> 
> -----Original Message-----
> From: glpk xypron [mailto:address@hidden 
> Sent: Thursday, July 22, 2010 3:25 PM
> To: Chelaru, Laura; address@hidden
> Subject: Re: Floating point precision?
> 
> Hello Laura,
> 
> the glpsol online helps says:
>   Options specific to simplex solver:
>    --exact           use simplex method based on exact arithmetic
> 
> The exact option is ignored for MIPs
> 
> > subject to Bin_d1 {i in I1}: d[i] <= 100*BIN2[i];
> > subject to Bin_d2 {i in I1}: d[i] >= k - 100*(1-BIN2[i]);
> Check the value range of d[i]. Could a smaller number then
> 100 be used?
> 
> Best regards
> 
> Xypron
> 
> 
> -------- Original-Nachricht --------
> > Datum: Thu, 22 Jul 2010 10:58:44 -0400
> > Betreff: Floating point precision?
> 
> > Hi,
> > 
> > I'm kind of new at this, so please bear with me. I am running a MIP
> using
> > the GLPK standalone solver. One of the constraints in my program is that
> a
> > set of variables have to be greater than 'k' if they are positive. I've
> > noticed that a couple of the variables fall within the [0, k] interval,
> and
> > I'm not sure why. I have specified using exact arithmetic.
> > 
> > These are the constraints I'm using:
> > subject to Bin_d1 {i in I1}: d[i] <= 100*BIN2[i];
> > subject to Bin_d2 {i in I1}: d[i] >= k - 100*(1-BIN2[i]);
> > 
> > where var BIN2 {i in I3} binary;
> > and k is around 0.01;
> > 
> > Thanks for your help,
> > Laura
> > 
> 
> -- 
> GMX DSL: Internet-, Telefon- und Handy-Flat ab 19,99 EUR/mtl.  
> Bis zu 150 EUR Startguthaben inklusive! http://portal.gmx.net/de/go/dsl
> 
> ********************
> NOTICE OF CONFIDENTIALITY
> This communication including any information transmitted with it is 
> intended only for the use of the addressees and is confidential. 
> If you are not an intended recipient or responsible for delivering 
> the message to an intended recipient, any review, disclosure, 
> conversion to hard copy, dissemination, reproduction or other use 
> of any part of this communication is strictly prohibited, as is the 
> taking or omitting of any action in reliance upon this communication. 
> If you receive this communication in error or without authorization 
> please notify us immediately by return e-mail or otherwise and 
> permanently delete the entire communication from any computer, 
> disk drive, or other storage medium.
> 
> If the above disclaimer is not properly readable, it can be found at 
> www.td.com/legal
>                                                            
> AVERTISSEMENT DE CONFIDENTIALITE                   
> Ce courriel, ainsi que tout renseignement ci-inclus, destiné uniquement 
> aux destinataires susmentionnés,  est confidentiel.  Si vous 
> n'êtes pas le destinataire prévu ou un agent responsable de la 
> livraison de ce courriel, tout examen, divulgation, copie, impression, 
> reproduction, distribution, ou autre utilisation d'une partie de ce 
> courriel est strictement interdit de même que toute intervention ou 
> abstraction à cet égard.  Si vous avez reçu ce message par erreur ou 
> sans autorisation, veuillez en aviser immédiatement l'expéditeur par 
> retour de courriel ou par un autre moyen et supprimer immédiatement 
> cette communication entière de tout système électronique.
> 
> Si l'avis de non-responsabilité ci-dessus n'est pas lisible, vous 
> pouvez le consulter à www.td.com/francais/legale

-- 
Neu: GMX De-Mail - Einfach wie E-Mail, sicher wie ein Brief!  
Jetzt De-Mail-Adresse reservieren: http://portal.gmx.net/de/go/demail