Towards a Network Interpretation of Agent Interaction in Ant Colony Optimization

This work introduces a novel framework for a network interpretation of agent interaction in ant-inspired algorithms. A complex network interpretation of population dynamics is a recent trend in the research of population-based metaheuristic algorithms. Complex network models of nature-inspired methods enable the use of a wide variety of analytical methods from the areas of graph theory and network science in the field of computational intelligence. Agent interaction is in this approach cast into an evolving complex network with vertices representing individual agents and arcs with evolving weights corresponding to their interaction. This paper presents a generic framework for such network interpretation of ant interaction as well as its initial implementation for a sample problem, the travelling salesman problem. Initial computational experiments illustrate the proposed concepts and demonstrate the usefulness of network-based analysis of ant-inspired methods.