Abstract � Adaptation of parameters and operators is one of the most important and promising areas of research in evolutionary computation � it tunes the algorithm to the problem while solving the problem. In this paper we develop a classi�cation of adaptation on the basis of the mechanisms used � and the level at which adaptation operates within the evolutionary algorithm. The classi�cation covers all forms of adaptation in evolutionary computation and suggests fur� ther research. I.

Abstract. In this paper we evaluate on-the-fly population (re)sizing mechanisms for evolutionary algorithms (EAs). Evaluation is done by an experimental comparison, where the contestants are various existing methods and a new mechanism, introduced here. These comparisons consider EA performance in terms of success rate, speed, and solution quality, measured on a variety of fitness landscapes. These landscapes are created by a generator that allows for gradual tuning of their characteristics. Our test suite covers a wide span of landscapes ranging from a smooth one-peak landscape to a rugged 1000-peak one. The experiments show that the population (re)sizing mechanisms exhibit significant differences in speed, measured by the number of fitness evaluations to a solution and the best EAs with adaptive population resizing outperform the traditional genetic algorithm (GA) by a large margin. 1

In this study, we give a brief overview of search strategy adaptation in evolutionary computation. The

This paper reviews the topic of population sizing in genetic algorithms. It starts by revisiting theoretical models which rely on a facetwise decomposition of genetic algorithms, and then moves on to various self-adjusting population sizing schemes that have been proposed in the literature. The paper ends with recommendations for those who design and compare adaptive population sizing schemes for genetic algorithms.

Evolutionary computation techniques have received a lot of attention regarding their potential as optimization techniques for complex real-world problems. These techniques, based on the powerful principle of "survival of the fittest", model some natural phenomena of genetic inheritance and Darwinian strife for survival; they also constitute an interesting category of modern heuristic search. This introductory article presents the main paradigms of evolutionary algorithms (genetic algorithms, evolution strategies, evolutionary programming, genetic programming) as well as other (hybrid) methods of evolutionary computation. Two particular research directions (parallel evolutionary techniques and self-adaptation) are discussed further in the last part of this paper. 1 Introduction The evolutionary computation (EC) techniques are stochastic algorithms whose search methods model some natural phenomena: genetic inheritance and Darwinian strife for survival. As stated in [33]: "... the metaph...

Evolutionary Algorithms are search algorithms based on the Darwinian metaphor of “Natural Selection”. Typically these algorithms maintain a population of individual solutions, each of which has a fitness attached to it, which in some way reflects the quality of the solution. The search proceeds via the iterative generation, evaluation and possible incorporation of new individuals based on the current population, using a number of parameterised genetic operators. In this thesis the phenomenon of Self Adaptation of the genetic operators is investigated. A new framework for classifying adaptive algorithms is proposed, based on the scope of the adaptation, and on the nature of the transition function guiding the search through the space of possible configurations of the algorithm. Mechanisms are investigated for achieving the self adaptation of recombination and mutation operators within a genetic algorithm, and means of combining them are investigated. These are shown to produce significantly better results than any of the combinations of fixed operators tested, across a range of problem types. These new operators reduce the need for the designer of an algorithm to select

The paper investigates a new Patchwork model for structured population in evolutionary search, where population size may vary. This model allows better control of both population diversity and selective pressure, and its operators are local in scope. Moreover, the Patchwork model gives a signicant exibility for introducing many additional concepts, like behavioral rules for individuals. First experiments allowed us to observe some interesting patterns which emerged during evolutionary process.

In this paper we address an extension of a very efficient genetic algorithm (GA) known as Hy3, a physical parallelization of the gradual distributed real-coded GA (GD-RCGA). This search model relies on a set of eight subpopulations residing in a cube topology having two faces for promoting exploration and exploitation. The resulting technique has been shown to yield very accurate results in continuous optimization by using crossover operators tuned to explore and exploit the solutions inside each subpopulation. We introduce here a further extension of Hy3, called Hy4, that uses 16 islands arranged in a hypercube of four dimensions. Thus, two new faces with different exploration/exploitation search capabilities are added to the search performed by Hy3. We analyze the importance of running a synchronous versus an asynchronous version of the models considered. The results indicate that the proposed Hy4 model overcomes the Hy3 performance because of its improved balance between exploration and exploitation that enhances the search. Finally, we also show that the async Hy4 model scales better than the sync one.

Exploration efficiency of GAs depends on parameter values such as mutation rate and crossover rate. To save labor of manually adjusting these values, GAs which automatically adjust parameter values(Adaptive GAs) have been proposed. We have proposed Self Adaptive Island GA(SAIGA), which does not require adjusting parameter values, and has search performance comparable to that of SGA with manually optimized parameter values. There are also parallel adaptive GAs proposed, but existing parallel adaptive GAs require periodical synchronization, and thus performance degrades if there are differences in performance between computers executed in parallel. In this paper, we propose Asynchronous SAIGA which does not require synchronization. We confirmed that it has same or more search performance than SAIGA.

A vast number of very successful applications of Memetic algorithms (MAs) have been reported in the literature in the last years for a wide range of problem domains. The majority of the papers dealing with MAs are the result of the combination of highly specialised preexisting local searchers and usually purpose-speci c genetic operators. Moreover, those algorithms require a considerable eort devoted to the tuning of the local search and evolutionary parts of the algorithm. We have demonstrated in our previous work (see references below), that given a range of possible local search strategies available to a Memetic Algorithm, the optimal choice of which one must be used is not only problem and instance dependent but also tightly related to the state of the search process itself. We also showed that it is indeed possible to produce Memetic Algorithms that adapt on-the-y to those situations for a variety of problem domains. In this paper we continue our studies of the design of robust Memetic Algorithms by introducing the concept of \self-generating" Memetic Algorithms. As mentioned above the success of a Memetic Algorithm depends on the pre-existence of powerful local searchers. Here we allow the Memetic Algorithm to create its local searchers and to co-evolve the behaviours it needs to successfully solve a problem. 1