I have a general question about column generation for large-scale systems. I understand the Master problem and the Restricted master problem for the subset of variables. After I set up the RMP, I find the duals of the system. I then set up the pricing problem and find the minimum reduced cost to determine next entering variable into the system.
My question is about the subsets of variables. The pricing problem is only selecting the next column from the subset of variables, at what point is the subset modified to cover the other variables? My problem has roughly 114 million decision variables, with 113 binary. In my initial subset I include about 230,000 DVs and can iterate through that fairly quickly.
Thanks in advance,
Please sign in to leave a comment.