Υβριδικοί εξελικτικοί αλγόριθμοι βελτιστοποίησης και εφαρμογές σε προβλήματα συνδυαστικής βελτιστοποίησης
Πέτικας, Ισίδωρος Α.
The objective of this thesis is the study and research of effective algorithmic approaches for addressing computationally hard optimization problems, through the analysis and the proposal of hybrid evolutionary methods and the formulation of the framework for applying them to specific problems. Initially, the mechanisms that rule the evolutionary methods and other metaheuristics (in particular those of the local search algorithms) are investigated and thoroughly analyzed. In addition, an extensive survey of their implementation on classical and modern combinational optimization problems is performed, with an emphasis on their hybrid schemes. The main goal of the hybrid evolutionary approach that is analyzed and proposed is to exploit the separate advantages and to encounter the weaknesses of the evolutionary processes and the local search algorithms, by combining them in a unified scheme. The approach consists of two stages: In the first stage, the evolutionary process is used for global exploration of the search space, while in the second stage a local search method is employed in order to take advantage of the knowledge obtained from global exploration. A genetic algorithm and a generalized pattern search algorithm are selected respectively for the implementation and the evaluation of the approach. Depending on the nature and the formulation of the optimization problem, distinction between the restrictive conditions with respect to their importance is introduced and different treatment of the candidate solutions that violate each category of constraints is performed.