On Wed, 2013-03-06 at 14:23 -0600, David T. Price wrote:
> The test for dual infeasibility is wrong. 
> It does not have a test for 
> rows or columns of type GLP_DB (at least when lb < ub). See glpapi12.c, 
> lines 720-740.
> 
> 

I don't think so. In glpk dual variables (i.e. reduced costs) for
double-bounded primal variables are combined Lagrange multipliers, which
may have any sign in the dual feasible solution (like for fixed primal
variables), so such dual variables need not to be tested.

For details please see Section 4.1 "Background", in particular, the
table on p.118, in the glpk reference manual.


Andrew Makhorin


_______________________________________________
Bug-glpk mailing list
[email protected]
https://lists.gnu.org/mailman/listinfo/bug-glpk

Reply via email to