[Help-glpk] Dual infeasibility

[Anand Seshadri]

Hi,
I have a large LP which I am trying to solve. GLPK
gives the output like this

Karush-Kuhn-Tucker optimality conditions:

KKT.PE: max.abs.err. = 2.25e-011 on row 281
        max.rel.err. = 3.66e-014 on row 85124
        High quality

KKT.PB: max.abs.err. = 9.56e-006 on row 33251
        max.rel.err. = 1.73e-007 on row 33251
        Medium quality

KKT.DE: max.abs.err. = 7.41e-009 on column 47780
        max.rel.err. = 1.28e-013 on column 34867
        High quality

KKT.DB: max.abs.err. = 1.04e+006 on column 14390
        max.rel.err. = 2.10e+003 on column 25697
        DUAL SOLUTION IS INFEASIBLE

End of output

Also the infeasibility is very small in the primal
problem (1e-12). I thought that the primal BFS need
not be Dual feasible unless it is optimal. So why does
GLPK need dual feasibility before the problem has run
to optimallity? Can anyone help me in here

Thanks,
Anand.


      
____________________________________________________________________________________
Looking for last minute shopping deals?  
Find them fast with Yahoo! Search.  
http://tools.search.yahoo.com/newsearch/category.php?category=shopping