Bedau et al.’s statistical classification system for long-term evolutionary dynamics provides a test for open-ended evolution. Making this test more rigorous, and passing it, are two of the most important open problems for research into systems of agents that actively process and transform their environment. This paper presents a detailed description of the application of this test to ‘Geb’, a system designed to verify and extend theories behind the generation of evolutionarily emergent systems. The result is that, according to these statistics, Geb exhibits unbounded evolutionary dynamics, making it the first autonomous artificial system to pass this test. However, having passed it, the most prudent course of action is to look for weaknesses in the test. The test is criticized, most significantly with regard to its normalization method for artificial systems. Furthermore, this paper presents a modified normalization method, based on component activity normalization, that overcomes these criticisms. The results of the revised test, when applied to Geb, indicate that this system does indeed exhibit open-ended evolution.

We mowen nat, although we hadden it sworn, It overtake, it slit awey so faste. It wole us maken beggers atte laste! Evolution is notorious for its creative power, but also for giving rise to complex, unpredictable dynamics. As a result, practitioners of artificial evolution have encountered difficulties in predicting, analysing, or even understanding the outcome of their experiments. In particular, the concept of evolutionary “progress ” (whether in the sense of performance increase or complexity growth) has given rise to much debate and confusion. After a careful description of the mechanisms of evolution and natural selection, we provide usable concepts of performance and progress in coevolution. In particular, we introduce a distinction between three types of progress: local, historical, and global, which we suggest underlies much of the confusion that surrounds coevolutionary dynamics. Similarly, we provide a comprehensive answer to the question of whether an “arrow of complexity ” exists in evolution. We introduce several methods to detect and analyse performance and progress in coevolutionary experiments. We propose a statistical

We introduce fitness transmission as a simple statistical signature of adaptive evolution within a system. Fitness transmission is the correlation between the fitness of parents and children, where fitness is evaluated after the number of grandchildren, suitably normalised. This measure is a direct calculation based on a genealogical record, rather than on genetic or phenotypic observation. We point out that the Bedau-Packard statistics of evolutionary activity cannot be used as a reliable system-wide signature of adaptive evolution, because they can produce positive signals when applied to certain “random”, non-evolutionary systems. We apply fitness transmission to simple evolutionary algorithms (as well as neutral equivalents) and demonstrate its capacity to accurately detect the presence or absence of Darwinian evolution.

We introduce differential fitness transmission as a signature of adaptive, Darwinian evolution, which can be detected using genealogical records of a reproducing population. This method is motivated by the observation that Darwinian evolution fundamentally consists in the differential transmission of heritable, fitness-affecting traits that result in differential transmission of fitness itself: fitter parents should tend to produce fitter offspring. Based on this idea, we propose several statistics which allow us to detect the presence of differential fitness transmission under various conditions. As an experimental illustration, we apply our statistics to simple evolutionary algorithms using different selection and replacement regimes. We demonstrate that differential fitness transmission can be used to detect the presence of a force which consistently favours certain lineages at the expense of others, over more than one generation. We conclude that differential fitness transmission is a useful, practical signature of Darwinian evolution for situations in which a genealogical record of the population can be obtained.

An experimental analysis of evolutionary algorithms usually generates a huge amount of multidimensional data, including numeric and symbolic data. It is difficult to efficiently navigate in such a set of data, for instance to be able to tune the parameters or evaluate the efficiency of some operators. Usual features of existing EA visualisation systems consist in visualising time- or generation-dependent curves (fitness, diversity, or other statistics). When dealing with genomic information, the task becomes even more difficult, as a convenient visualisation strongly depends on the considered fitness landscape. In this latter case the raw data are usually sets of successive populations of points of a complex multidimensional space. The purpose of this paper is to evaluate the potential interest of a recent visual analytics tool for navigating in complex sets of EA data, and to sketch future developements of this tool, in order to better adapt it to the needs of EA experimental analysis.

apport de recherche