In this paper, we develop techniques based on evolvability statistics of the fitness landscape surrounding sampled solutions. Averaging the measures over a sample of equal fitness solutions allows us to build up fitness evolvability portraits of the fitness landscape, which we show can be used to compare both the ruggedness and neutrality in a set of tunably rugged and tunably neutral landscapes. We further show that the techniques can be used with solution samples collected through both random sampling of the landscapes and online sampling during optimization. Finally, we apply the techniques to two real evolutionary electronics search spaces and highlight differences between the two search spaces, comparing with the time taken to find good solutions through search.

Abstract—Self-adaptation is used in all main paradigms of evolutionary computation to increase efficiency. We claim that the basis of self-adaptation is the use of neutrality. In the absence of external control neutrality allows a variation of the search distribution without the risk of fitness loss. I.

This paper investigates how the use of redundant representations influences the performance of genetic and evolutionary algorithms. Representations are redundant if the number of genotypes exceeds the number of phenotypes. A distinction is made between synonymously and nonsynonymously redundant representations. Representation are synonymously redundant if the genotypes that represent the same phenotype are very similar to each other. Non-synonymously redundant representations do not allow genetic operators to work properly and result in a lower performance of evolutionary search. When using synonymously redundant representations, the performance of selectorecombinative genetic algorithms (GAs) depends on the modification of the initial supply. Theoretical models are developed that show the necessary population size to solve a problem and the number of generations goes with O(2 /r), where k r is the order of redundancy and r is the number of genotypic building blocks (BB) that represent the optimal phenotypic BB. Therefore, uniformly redundant representations do not change the behavior of GAs. Only by increasing r, which means overrepresenting the optimal solution, does GA performance increase. Therefore, non-uniformly redundant representations can only be used advantageously if a-priori information exists regarding the optimal solution. The validity of the proposed theoretical concepts is illustrated for the binary trivial voting mapping and the realvalued link-biased encoding. The empirical investigations show that the developed population sizing and time to convergence models allow an accurate prediction of the empirical results.

In this paper, we investigate a neutral epoch during an optimisation run with complex genotype-to-fitness mapping. The behaviour of the search process during neutral epochs is of importance for evolutionary robotics and other artificial-life approaches that evolve problem solutions; recent work has argued that evolvability may change during these epochs. We investigate the distribution of offspring fitnesses from the best individuals of each generation in a population-based genetic algorithm, and see no trends towards higher probabilities of producing higher fitness offspring, and no trends towards higher probabilities of not producing lower fitness offspring. A second experiment in which populations from across the neutral epoch are used as initial populations for the genetic algorithm, shows no difference between the populations in the number of generations required to produce high fitness. We conclude that there is no evidence for change in evolvability during the neutral epoch in this optimisation run; the population is not doing anything “useful” during this period.

Evolutionary algorithms apply the process of variation, reproduction and selection to look for an individual capable of solving the task at hand. In order to improve the evolvability of a population we propose to copy important characteristics of nature's search space. Desired characteristics for a genotype-phenotype mapping are described and several highly redundant genotype-phenotype mappings are analyzed in the context of a population based search. We show that evolvability, de ned as the ability of random variations to sometimes produce improvement, is inuenced by the existence of neutral networks in genotype space. Redundant mappings allow the population to spread along the network of neutral mutations and the population is quickly able to recover after a change has occurred. The extent of the neutral networks aects the interconnectivity of the search space and thereby aects evolvability.

We propose building neutral networks in needle-in-haystack fitness landscapes to assist an evolutionary algorithm to perform search. The experimental results on four different problems show that this approach improves the search success rates in most cases. In situations where neutral networks do not give performance improvement, no impairment occurs either.

Abstract. Neutral genotype-phenotype mappings can be observed in natural evolution and are often used in evolutionary computation. In this article, important aspects of such encodings are analyzed. First, it is shown that in the absence of external control neutrality allows a variation of the search distribution independent of phenotypic changes. In particular, neutrality is necessary for self-adaptation, which is used in a variety of algorithms from all main paradigms of evolutionary computation to increase efficiency. Second, the average number of fitness evaluations needed to find a desirable (e.g., optimally adapted) genotype depending on the number of desirable genotypes and the cardinality of the genotype space is derived. It turns out that this number increases only marginally when neutrality is added to an encoding presuming that the fraction of desirable genotypes stays constant and that the number of these genotypes is not too small.

Abstract. The effects of neutrality on evolutionary search are not fully understood. In this paper we make an effort to shed some light on how and why bit-wise neutrality – an important form of neutrality induced by a genotype-phenotype map where each phenotypic bit is obtained by transforming a group of genotypic bits via an encoding function – influences the behaviour of a mutation-based GA on functions of unitation. To do so we study how the fitness distance correlation (fdc) of landscapes changes under the effect of different (neutral) encodings. We also study how phenotypic mutation rates change as a function of the genotypic mutation rate for different encodings. This allows us to formulate simple explanations for why the behaviour of a GA changes so radically with different types of neutrality and mutation rates. Finally, we corroborate these conjectures with extensive empirical experimentation. 1

In natural systems, animals discriminate an object through information coming from various receptors. In particular, object’s figure and its motion pattern are known to be very important for quick and accurate discrimination. In this work, to give the similar ability of discrimination to an artificial agent, we examine whether artificial evolution is capable of generating artificial neural networks that perform discrimination tasks using the mixed information of figure and motion pattern. The results demonstrate that evolutionary approach is successful in developing the neural network controller using a affordable computational cost. 1

This paper investigates how the use of the trivial voting (TV) mapping influences the performance of genetic algorithms (GAs). The TV mapping is a redundant representation for binary phenotypes. A population sizing model is presented that quantitatively predicts the influence of the TV mapping and variants of this encoding on the performance of GAs. The results indicate that when using this encoding GA performance depends on the influence of the representation on the initial supply of building blocks. Therefore, GA performance remains unchanged if the TV mapping is uniformly redundant that means on average a phenotype is represented by the same number of genotypes. If the optimal solution is overrepresented, GA performance increases, whereas it decreases if the optimal solution is underrepresented. The results show that redundant representations like the TV mapping do not increase GA performance in general, but higher performance can only be achieved if there is specific knowledge about the structure of the optimal solution which can beneficially be used by the redundant representation.