Cooperative Coevolution: An Architecture for Evolving Coadapted Subcomponents
Evolutionary Computation
| Hi-index | 0.00 |
This paper offers a novel approach to coevolution based on the sociological theory of symbolic interactionism. It provides a multi-agent computational model along with experimental results that suggest improved fitness, robustness, and knowledge due to emergent symbol systems. The main contribution of the symbolic-interactionist approach to coevolution is the concept of the emergence of a system in the abstract, where an interface between agents evolves. The interface is an emergent symbol system that focuses selective pressure among agents in ways that have been beneficial to agents as a whole in the past, creating a coevolving system that takes advantage of epistasis rather than having to prevent it. Global fitness thereby emerges from local, selfish interaction. The assignment of roles in this system is endogenous.