Evolvable Agents in Static and Dynamic Optimization Problems
Methods
| Title | Evolvable Agents in Static and Dynamic Optimization Problems |
| Publication Type | Conference Paper |
| Year of Publication | 2008 |
| Authors | Laredo, Juan L., Castillo Pedro A., Mora Antonio M., Merelo Juan J., Rosa Agostinho, and Fernandes Carlos |
| Conference Name | Proceedings of the 10th international conference on Parallel Problem Solving from Nature |
| Pagination | 488-497 |
| Publisher | Springer-Verlag |
| Conference Location | Berlin, Heidelberg |
| ISBN Number | 978-3-540-87699-1 |
| Abstract | This paper investigates the behaviour of the Evolvable Agent model (EvAg) in static and dynamic environments. The EvAg is a spatially structured Genetic Algorithm (GA) designed to work on Peer-to-Peer (P2P) systems in which the population structure is a small-world graph built by newscast, a P2P protocol. Additionally to the profits in computing performance, EvAg maintains genetic diversity at the small world relationships between individuals in a sort of social network. Experiments were conducted in order to assess how EvAg scales on deceptive and non-deceptive trap functions. In addition, the proposal was tested on dynamic environments. The results show that the EvAg scales and adapts better to dynamic environments than a standard GA and an improved version of the well-known Random Immigrants Genetic Algorithm. |
| URL | http://dx.doi.org/10.1007/978-3-540-87700-4_49 |
| DOI | 10.1007/978-3-540-87700-4_49 |
| Citation Key | Laredo2008 |
- Login to post comments
