assume someone implements branch-and-cut... and needs to solve the LP relaxations in different nodes. This means, after solving the LP of one node, we can switch to a very different node (different constraints and variable bounds) and reusing the basis is not efficient.
My approach to this would be to store basis information in each node and use it to warmstart the child's LP, but this appears to fail because Gurobi requires a complete and correct basis. What is the right approach? Solve the parent's LP with the stored basis, then modify the problem according to the child and resolve?
Please sign in to leave a comment.