Self-configurable memetic algorithm
Le, Minh Nghia
Date of Issue2012
School of Computer Engineering
Centre for Computational Intelligence
Honda Research Institute Europe, Germany
To date, most successful advanced stochastic optimization algorithms involve some forms of individual learning or meme in their design. Memetic Algorithm (MA), as a form of hybridization between population-based and individual-based searches, represents one of the recent growing areas in evolutionary algorithm research. In the success and surge in interests on MAs, researchers have been exploring on various possible hybridizations of search operators towards the development and manual crafting of specialized algorithms that solve a specific problem or a set of problems effectively, using the domain knowledge obtained from human expertise. However, with so many population-based and individual-based procedures available for hybridizing, it is a tedious task, if not impossible, to design in advance an effective memetic algorithm for a given problem at hand. Furthermore, when high-fidelity analysis codes are used for evaluating design solutions in the optimization process, it is not uncommon for the single simulation process to take minutes, hours to days of supercomputer time to compute. Since the design cycle time of a product is directly proportional to the number of calls made to the costly analysis solvers, there has been practical needs for novel meta-model/surrogate-assisted memetic frameworks that can handle these forms of problems elegantly.
DRNTU::Engineering::Computer science and engineering::Theory of computation::Analysis of algorithms and problem complexity