Hybrid Metaheuristic Algorithms and their Applications to Face Recognition

Abstract

The thesis presents novel hybrid metaheuristic algorithms and their applications to the face recognition. Many metaheuristic algorithms are developed during the last three decades, and applied successfully in various optimization problems. The future lies in the hybridization of metaheuristic algorithms, where two or more metaheuristic algorithms search process are combined to form a hybrid metaheuristic algorithm. This thesis presents three new hybrid metaheuristic algorithms known as crossover bacterial foraging optimization algorithm (CBFOA), adaptive CBFOA (ACBFOA), and adaptive cuckoo search (ACS) algorithm. The face recognition problem is interesting yet challenging. After the development of these hybrid metaheuristic algorithms, we implement them in the development of face recognition methods known as CBFO-Fisher, ACBFO-Fisher, and ACS-IDA. The CBFOA is a hybrid metaheuristic algorithm developed by hybridizing the crossover operator of the genetic algorithm (GA) in the reproduction steps of a bacterial foraging optimization algorithm (BFOA). The idea of using the crossover mechanism is to search nearby locations by offspring (50 percent of bacteria), because they are randomly produced at different locations, and other (50 percent) bacteria are retained by the BFOA itself. This helps the CBFOA to handle the exploration of the search space more efficiently. Seven benchmark test functions are used for the performance evaluation. The comparison of the CBFOA with the previous algorithms reveals its effectiveness. newline