How Different Hierarchical Relationships Impact Evolution

摘要

The evolutionary behavior of three hierarchical relationships, HIFF-C, HIIFF-II and HIFF-M is studied in the context of two computational models, J and JGA. In J, entities are composed from other entities in the population. JGA is a panmictic genetic algorithm. Results from our experiments indicate that specificity in a relationship enhances convergence to a global optimum in both models. When there is little specificity in the relationship, external conditions such as join rate, crossover rate, agitation type or selection mechanism need to be set appropriately. Our results also suggest that cooperation is neither necessary nor sufficient for the evolution of higher level entities. We found that cooperation was evolutionary advantages in J only for relationships with little to no top-down inter-level conflict.