Originally posted by: SystemAdmin

[bistra said:]

I am writing a IloCplex::BranchCallbackI and I get to a node where I have an integer boolean variable that is set to an integer value but its current bounds are infeasible, i.e. getLB(var) > getUB(var)

Is it safe to prune this node as infeasible?

thank you,
bistra
#CPLEXOptimizers
#DecisionOptimization

Originally posted by: SystemAdmin

[bistra said:]

I found a similar topic posted some time back:
"Inconsistent bounds while using a Branching Callback"

[url=http://forums.ilog.com/optimization/index.php?topic=106.0]http://forums.ilog.com/optimization/index.php?topic=106.0[/url]

However, the replies to this post just say that it is unclear what cplex has done in its own variable selection.
I am using strong branching (variable selection 3) as the default branching selection (which is pretty strong) but sometimes I want to override the branch variable selected.

I am still not clear whether it is safe to conclude that this node and any possible children of it are infeasible because of the current inconsistent bounds...
#CPLEXOptimizers
#DecisionOptimization

Originally posted by: SystemAdmin

[Sylvain said:]

My experience is that it is not safe to prune such nodes.

My understanding was that branching on variables which have a feasibility status "ImpliedFeasible" would generate these inconsistent bounds.

I just think it is some sort of bug which does not affect the result of the optimization but makes life more difficult for the user who develops BranchCallbacks.
#CPLEXOptimizers
#DecisionOptimization