Abstract | ||
---|---|---|
We present a novel visualisation method for island-based evolutionary algorithms based on the concept of tracers as adopted in medicine and molecular biology to follow a biochemical process. For example, a radioisotope or dye can be used to replace a stable component of a biological compound, and the signal from the radioisotope can be monitored as it passes through the body to measure the compound's distribution and elimination from the system. In a similar fashion we attach a tracer dye to individuals in each island, where each individual in any one island is marked with the same colour, and each island then has its own unique colour signal. We can then monitor how individuals undergoing migration events are distributed throughout the entire island ecosystem, thereby allowing the user to visually monitor takeover times and the resulting loss of diversity. This is achieved by visualising each island as a spectrum of the tracer dye associated with each individual. Experiments adopting different rates of migration and network connectivity confirm earlier research which predicts that island models are extremely sensitive to the size and frequency of migrations. |
Year | DOI | Venue |
---|---|---|
2011 | 10.1007/s10710-010-9125-y | Genetic Programming and Evolvable Machines |
Keywords | Field | DocType |
Visualisation,Distributed evolutionary computation | Data mining,Network connectivity,TRACER,Evolutionary algorithm,Computer science,Visualization,Evolutionary computation,Artificial intelligence,Machine learning | Journal |
Volume | Issue | ISSN |
12 | 2 | 1389-2576 |
Citations | PageRank | References |
1 | 0.39 | 14 |
Authors | ||
3 |
Name | Order | Citations | PageRank |
---|---|---|---|
Michael O'Neill | 1 | 876 | 69.58 |
Anthony Brabazon | 2 | 918 | 98.60 |
Erik Hemberg | 3 | 143 | 35.68 |