The Evolution of Complexity - Abstracts.

Stepping up Trophic Levels with Self-Repairing Genetic Algorithm

By Yukihiko Toquenaga

Abstract:

We modified our artificial ecosystem (Saruwatari et al. 1994) to investigate the relationship among species diversity, trophic interaction, and system persistency in a homogeneous spatial structure. Artificial creatures lived in finite number of cells and there was no spatial structure. The system was continuously showered by sun light. Individual creature was represented by a character bit-string. We adopted a new genetic algorithm based on the hypothesis that gene repairing is an important driving force in evolution (Avise 1993). In stead of encoding mutation rates into the gene code, we allowed the genetic strings to accommodate the exogenous severe mutation force by folding themselves into two and repairing damages. The two arms of a folded string mutually complemented the bits destroyed by mutation. The phenotype of each individual was determined by the bit pattern of the longer arm of its folded string. At each time step, an individual tried to consume both the sun light and another individual in the world. It obtained as much energy as the Hamming distance between its own phenotype and its victimIT. Larger individuals had benefit to use the most profitable part of their bit-strings at the expense of maintenance cost proportional to the phenotype length.

In contrast to our previous study, the present ecosystem showed relatively high persistency. The results were classified into the following six categories; 1) Simple Extinction: The initial populations could not establish. This situation occurred when the world size was small and energy efficiency was low. 2) Freezing: The world was occupied by small homogeneous individuals. This situation occurred when the world size was small. 3) Calm Stepping-up: The individual phenotype showed punctuating elongation and most cells were continuously occupied. The maximum body sizes were about 15 and 8 in the gene-repairing and unrepairing systems, respectively. 4) Moderate Stepping-up: The individual phenotype showed punctuating, but the population size began to fluctuate. Only 80 % cells were occupied in the later time steps. This situation only occurred in gene-repairing systems when the world size was large and energy efficiency was high. 5) Chaotic Stepping-up: The individual phenotype showed punctuating elongation, but the population size began to drastically fluctuate. Only 50 % cells were occupied in the later time steps. This situation occurred when the world size was large and the energy efficiency was low. 6) Long-run Extinction: The system crashed after long-run and the maximum phenotype size was about 10. This situation occurred when the world size was small and energy efficiency was low. The relative high persistency in our new model was achieved by new feeding interaction rule rather than by gene-repairing GA. However, the equilibrium of phenotype size was much higher in the system with gene-repairing than those without repairing. Genetic analysis showed that population consisted of two types of individuals: highly variable (without turning points) and conservative ones (capable with self-repairing).