Genetic algorithms (GAs) and simulated annealing (SA) are two important search methods that have been used successfully in solving difficult problems such as combinatorial optimization problems. Genetic algorithms are capable of wide exploration of the search space, while simulated annealing is capable of fine tuning a good solution. Combining both techniques may result in achieving the benefits of both and improving the quality of the solutions obtained. Several attempts have been made to hybridize GAs and SA. One such attempt was to augment a standard GA with simulated annealing as a genetic operator. SA in that case acted as a directed or intelligent mutation operator as opposed to the random, undirected mutation operator of GAs. Although using this technique showed some advantages over GA used alone, one problem was to find fixed global annealing parameters that work for all solutions and all stages in the search process. Failing to find optimum annealing parameters affects the quality of the solution obtained and may degrade performance. In this research, we try to overcome this weakness by introducing an adaptive hybrid GA - SA algorithm, in which simulated annealing acts as a special case of mutation. However, the annealing operator used in this technique is adaptive in the sense that the annealing parameters are evolved and optimized according to the requirements of the search process. Adaptation is expected to help guide the search towards optimum solutions with minimum effort of parameter optimization. The algorithm is tested in solving an important NP-hard problem, which is the MAP (Maximum a-Posteriori) assignment problem on BBNs (Bayesian Belief Networks). The algorithm is also augmented with some problem specific information used to design a new GA crossover operator. The results obtained from testing the algorithm on several BBN graphs with large numbers of nodes and different network structures indicate that the adaptive hybrid algorithm provides an improvement of solution quality over that obtained by GA used alone and GA augmented with standard non-adaptive simulated annealing. Its effect, however, is more profound for problems with large numbers of nodes, which are difficult for GA alone to solve.
Computer Science & Engineering Department
MS in Computer Science
Date of Award
Online Submission Date
Committee Member 1
Committee Member 2
The author retains all rights with regard to copyright. The author certifies that written permission from the owner(s) of third-party copyrighted matter included in the thesis, dissertation, paper, or record of study has been obtained. The author further certifies that IRB approval has been obtained for this thesis, or that IRB approval is not necessary for this thesis. Insofar as this thesis, dissertation, paper, or record of study is an educational record as defined in the Family Educational Rights and Privacy Act (FERPA) (20 USC 1232g), the author has granted consent to disclosure of it to anyone who requests a copy. The author has granted the American University in Cairo or its agents a non-exclusive license to archive this thesis, dissertation, paper, or record of study, and to make it accessible, in whole or in part, in all forms of media, now or hereafter known.
Not necessary for this item
(2000).An adaptive hybrid genetic-annealing approach for solving the map problem on belief networks [Master’s thesis, the American University in Cairo]. AUC Knowledge Fountain.
Hosny, Manar. An adaptive hybrid genetic-annealing approach for solving the map problem on belief networks. 2000. American University in Cairo, Master's thesis. AUC Knowledge Fountain.