14h00 - 15h00
Recent Advances in Mixed-Integer Linear Programming
The research group at Carleton University has been pursuing a number of new directions in Mixed-Integer Linear Programmingin recent years. This talk summarizes our main new algorithms in two areas: branching methods for reaching integer feasibility faster, and node selection methods for reaching MILP optimality faster. We present a variety of new branching variable and branching
direction selection methods, new methods for general multi-variable branching disjunctions, and the unifying principle of branching to force change. We also present a number of node selection methods that exploit correlation and distribution characteristics of branch and bound trees to
trigger backtracking and to choose the next node to solve when backtracking.