Parameter Determination in Relax-and-Fix Heuristics
Relax-and-Fix (RF) heuristics are normally executed using parameters determined by trial-and-error. Using numerical methods, the parameters of the Relax-and-Fix heuristic can be optimized. In this thesis, the definition of a Relax-and-Fix parameter is reviewed. Then, the parameters are optimized using two different methods. At first, a method which seeks optimal subproblem orders for the RF heuristic is introduced. Secondly, the subproblem sizes of the RF heuristic are made dynamic, to check if the commonly used homogeneous subproblem size can be improved.