Multi-Population Genetic Programming with Data Migration for Symbolic Regression
M. Kommenda, M. Affenzeller, G. K. Kronberger, B. Burlacu, S. M. Winkler - Multi-Population Genetic Programming with Data Migration for Symbolic Regression in Computational Intelligence and Efficiency in Engineering Systems (Contributions to Book: Part/Chapter/Section 6), (Editors: G. Borowik, Z. Chaczko, L.G. Ford, W. Jacak, T. Luba) - Springer, 2015, pp. 75-87
In this contribution we study the effects of multi-population genetic programming for symbolic regression problems. In addition to the parallel evolution of several subpopulations according to an island model with unidirectional ring migration, the data partitions, on which the individuals are evolved, differ for every island and are adapted during algorithm execution. These modifications are intended to increase the generalization capabilities of the solutions and to maintain the genetic diversity. The effects of multiple populations as well as the used data migration strategy are compared to standard genetic programming algorithms on several symbolic regression benchmark problems.
- FH-Prof. PD DI Dr. Michael Affenzeller
- FH-Prof. DI Dr. Stephan Winkler
- FH-Prof. DI Dr. Gabriel Kronberger
- DI (FH) Michael Kommenda MSc
- Bogdan Burlacu