[Coin-discuss] GLPK AND SYMPHONY not getting the same solutions

[Ted Ralphs]

Yes, this is exactly the problem. According to the MPS standard, any
integer variable without bounds is assumed to be binary. The COIN MPS
reader complies with this standard, but not all readers do (GLPKs reader
does not, for instance).

Cheers,

Ted

Tobias Achterberg wrote:
> The problem seems to be in the MPS standard.
> 
> As you stated the MPS file, you defined ALL variables to be binary, because the MPS standard says
> that variables that are included in a
> 
>     MARK0000  'MARKER'                 'INTORG'
>     ...
>     MARK0001  'MARKER'                 'INTEND'
> 
> section are binary variables by default. If you want them to be general integer variables, you have
> to specify bounds for the variables in the BOUNDS section. That means, your
> 
> BOUNDS
>     BV    BND1   Y1
> 
> is redundant, but if you want to declare variables X1, X2, and X3 to be integer, you would have to
> include them in the BOUNDS section.
> 
> I attached a modified MPS file, in which the X variables are general integers without bounds (i.e.
> -inf <= Xi <= +inf). The solution to this problem is X1 = X2 = -299, X3 = Y = 0, obj = -5980000.
> 
> If the variables are defined to be general integer with a lower bound of 0, the solution is the zero
> vector.
> 
> Probably, there is a difference in the MPS file readers of Symphony and GLPK...
> 
> Regards, Tobi
> 
> 
> ZT wrote:
> 
>>Hello,
>>
>>im having problems getting the same solutions using
>>both programs. I attached the MPS file that containts
>>the problem, consisting of 3 integer variables and 1
>>binary. GLPK calculates the problem correctly,
>>SYMPHONY does not. Anyone knows how to get the same
>>solution out of SYMPHONY?
>>
>>thanks! bye
>>
>>
>>		
>>__________________________________ 
>>Yahoo! Music Unlimited 
>>Access over 1 million songs. Try it free.
>>http://music.yahoo.com/unlimited/
>>
>>
>>------------------------------------------------------------------------
>>
>>_______________________________________________
>>Coin-discuss mailing list
>>Coin-discuss at list.coin-or.org
>>http://list.coin-or.org/mailman/listinfo/coin-discuss
> 
> 
> 
> ------------------------------------------------------------------------
> 
> NAME          problem.lp
> ROWS
>  N  OBJECTIVE
>  L  BND00001 
>  L  BND00004 
>  L  BND00074 
>  L  BND00077 
>  L  BND00145 
>  L  BND00146 
>  L  BND00147 
>  L  BND00148 
>  L  BND00149 
>  L  BND00150 
>  L  BND00151 
>  L  BND00152 
>  L  BND00153 
>  L  BND00154 
>  L  BND00155 
>  L  BND00156 
>  L  BND00157 
>  L  BND00158 
>  L  BND00159 
>  L  BND00160 
>  L  BND00161 
>  L  BND00162 
>  L  BND00163 
>  L  BND00164 
>  E  BND00295 
>  E  BND00296 
> COLUMNS
>     MARK0000  'MARKER'                 'INTORG'
>     X1         OBJECTIVE         10000   BND00001              1
>     X1         BND00074              1   BND00145              1
>     X1         BND00147              1   BND00149              1
>     X1         BND00151         0.6667   BND00152         0.3333
>     X1         BND00153        -0.3333   BND00154        -0.6667
>     X1         BND00155        -0.3333   BND00156         0.3333
>     X1         BND00295             -1   BND00296             -1
>     X2         OBJECTIVE         10000   BND00004              1
>     X2         BND00077              1   BND00146              1
>     X2         BND00148              1   BND00150              1
>     X2         BND00157         0.6667   BND00158         0.3333
>     X2         BND00159        -0.3333   BND00160        -0.6667
>     X2         BND00161        -0.3333   BND00162         0.3333
>     X2         BND00295              1   BND00296              1
>     X3         OBJECTIVE          1000   BND00151         0.6667
>     X3         BND00152         0.3333   BND00153        -0.3333
>     X3         BND00154        -0.6667   BND00155        -0.3333
>     X3         BND00156         0.3333   BND00163             -1
>     X3         BND00164              1
>     Y1         BND00163            100   BND00164           -200
>     MARK0001  'MARKER'                 'INTEND'
> RHS
>     rhs        BND00001            180   BND00004            180
>     rhs        BND00074            180   BND00077            180
>     rhs        BND00145            180   BND00146            180
>     rhs        BND00147            180   BND00148            180
>     rhs        BND00149            180   BND00150            180
>     rhs        BND00151            200   BND00152            400
>     rhs        BND00153            700   BND00154            200
>     rhs        BND00155            400   BND00156            700
>     rhs        BND00157            200   BND00158            400
>     rhs        BND00159            700   BND00160            200
>     rhs        BND00161            400   BND00162            700
> BOUNDS
>  FR bnd        X1       
>  FR bnd        X2       
>  FR bnd        X3       
>  UP bnd        Y1                    1
> ENDATA
> 
> 
> ------------------------------------------------------------------------
> 
> _______________________________________________
> Coin-discuss mailing list
> Coin-discuss at list.coin-or.org
> http://list.coin-or.org/mailman/listinfo/coin-discuss


-- 
Dr. Ted Ralphs
Assistant Professor
Industrial and Systems Engineering
Lehigh University
(610)758-4784
tkralphs at lehigh.edu
www.lehigh.edu/~tkr2