[Coin-lpsolver] Relaxing Constraints with Integer Variables

[Lou Hafer]

Joseph,

	Can't answer specifically for clp, but a general comment: relaxations
of the form a'x >= x0 - Mb are in general loose and ugly to solve,
particularly when M is some arbitrary large value. If you are interested in
integral values for b (i.e., you're using b to model a disjunction), and
intend to solve more than one or two of these, it's well worth your time to
find the minimal value for M. Bound a'x using upper and lower bounds on the
x_j and use the bound to calculate M. Should take care of clp's complaint at
the same time.

	If these `big M' constraints are imbedded in a more general
branch-and-bound procedure, you might look at the CglProbing cut generator,
which will tighten bounds on variables.

							Lou