Re: [Help-glpk] Bug or confused user?

[glpk xypron]

Hello Reginald,

I guess the cause of your problem is a missing definiton of the bounds
of mc. 

The CPLEX LP file format has a default lower bound of 0 for
all variables without explicitely stated bounds.

For details refer to "C.4 Bounds section" of glpk-4.44/doc/glpk.pdf.

Take the following examples:

minimize
  z: + mc
subject to
  mc >= -3
bounds
  mc free
end

This will result in an objective value of -3.

minimize
  z: + mc
subject to
  mc >= -3
end

This will result in an objective value of 0.

Best regards

Xypron


-------- Original-Nachricht --------
> Datum: Wed, 1 Sep 2010 13:22:48 -0700 (PDT)
> CC: address@hidden
> Betreff: Re: [Help-glpk] Bug or confused user?

> xypron,
> 
> The example is a constrained, overdetermined L1 regression of a line, but
> it doesn't converge to the correct values for mc & m000_000.  Instead it
> terminates w/ mc=0 and m000_000 = 1.42372e+09 and a large residual error.  It
> should converge to mc = -2.5e4 & m000_000 = 1.8e9.
> 

-- 
GMX DSL SOMMER-SPECIAL: Surf & Phone Flat 16.000 für nur 19,99 €/mtl.!*
http://portal.gmx.net/de/go/dsl