Case-Based Initialization of Genetic Algorithms
Proceedings of the 5th International Conference on Genetic Algorithms
Performance Measures for Dynamic Environments
PPSN VII Proceedings of the 7th International Conference on Parallel Problem Solving from Nature
Cooperative Coevolution: An Architecture for Evolving Coadapted Subcomponents
Evolutionary Computation
Evolutionary Computation in Dynamic and Uncertain Environments (Studies in Computational Intelligence)
Cheating for problem solving: a genetic algorithm with social interactions
Proceedings of the 11th Annual conference on Genetic and evolutionary computation
Memory based on abstraction for dynamic fitness functions
Evo'08 Proceedings of the 2008 conference on Applications of evolutionary computing
Population-Based Incremental Learning With Associative Memory for Dynamic Environments
IEEE Transactions on Evolutionary Computation
| Hi-index | 0.00 |
Evolutionary algorithms were originally designed to locate basins of optimum solutions in a stationary environment. Therefore, additional techniques and modifications have been introduced to deal with further requirements such as handling dynamic fitness functions or finding multiple optima. In this paper, we present a new approach for building evolutionary algorithms that is based on concepts borrowed from social behaviour evolution. Algorithms built with the proposed paradigm operate on a population of individuals that move in the search space as they interact and form groups. The interaction follows a set of social behaviours evolved by each group to enhance its adaptation to the environment (and other groups) and to achieve different desirable goals such as finding multiple optima, maintaining diversity, or tracking a moving peak in a changing environment. Each group has two sets of behaviours: one for intra-group interactions and one for inter-group interactions. These behaviours are evolved using mathematical models from the field of evolutionary game theory. This paper describes the proposed paradigm and starts studying it characteristics by building a new evolutionary algorithm and studying its behavior. The algorithm has been tested using a benchmark problem generator with promising initial results, which are also reported.