[gecode-users] Pruning filtered branchers.

[Duane Leslie]

I'm not sure if this is a feature or a bug, but I have a problem where I am using the first brancher to implement iterative deepening with RBS.

I use a heuristic with the first brancher disable via the filter function to determine a 'reasonably good' solution to use for BAB pruning, and then once the bound has been determined I switch the brancher on to activate IDA.

I have found that if the filter function returns false the first time it is called it is never called again, so I have had to implement an initial harmless dummy action so that it looks like it has work to do.  I am assuming this is an optimisation that is pruning away the brancher because it assumes it will never have work to do if it has none initially, but is this the desired behaviour and if so is there a better way to get around it than having a dummy action?

Regards,

Duane.