(Apologies for crossposting I posted this first in the CPLEX folder but later realized there is a specialized subforum for Math Programming)

Hi All,

For long Bender's decomposition has been considered to be faster than straightforward solution of an MIP. Does anyone have any reference to recent work that shows that for relatively hard MIPs? I wrote a simple Bender's code for an MIP but straightforward CPLEX solution times were much better than Benders. So, I have my doubts.

Also, is there any plan to include Bender's decomposition as one of the options within CPLEX itself? Would it not be possible in the future to identify MIPs where, for instance, on fixing the IP variables, the rest of the problem becomes a Network Problem (say) and then CPLEX itself can solve the problem via Benders decomposition?

Thanks.