UCL School of Management is delighted to welcome Murat Koksalan, Ross, to host a research seminar discussing ‘Multiobjective Combinatorial Optimization: Representing Nondominated Points and Converging to Preferred Solutions’
Combinatorial optimization problems are computationally hard even for a single criterion. The problem becomes harder when multiple criteria are considered. There are many nondominated points that are candidates to be the most preferred point and there is a need to incorporate the decision maker’s (DM’s) preferences to converge. These issues bring additional difficulties both computationally and cognitively. There has been an increasing interest in multiobjective combinatorial optimization (MOCO) problems in recent years. In this talk, I will discuss characterizing the nondominated frontier by fitting a function to a few nondominated points. I will demonstrate the performance of this characterization on several MOCO problems and I will discuss an exact and an approximate interactive algorithm for finding the most preferred solution of a DM under a quasiconcave value function.