Related Experiment Videos
Dynamic subset selection based on a fitness case topology.
Christian W G Lasarczyk1, Peter Dittrich, Wolfgang Banzhaf
1Department of Computer Science, University Dortmund, Joseph-von-Fraunhofer-Str. 20, 44227 Dortmund, Germany. christian.lasarczyk@cs.uni-dortmund.de
Evolutionary Computation
|May 26, 2004
Summary
A novel topology-based subset selection method accelerates genetic programming by intelligently selecting fitness cases. This approach improves evolutionary search efficiency by considering problem structure without being problem-specific.
Area of Science:
- Computer Science
- Artificial Intelligence
- Evolutionary Computation
Background:
- Large fitness case sets significantly slow down genetic programming (GP).
- Existing subset selection methods may not fully leverage problem structure for efficiency.
Purpose of the Study:
- To introduce a new, problem-independent subset selection method for genetic programming.
- To improve the efficiency of evolutionary search by optimizing fitness case selection.
Main Methods:
- A topology is created on fitness cases based on their relationship, induced by individuals in the evolving population.
- The strength of the relation between fitness cases increases if an individual solves both.
- Topology-based subset selection prioritizes fitness cases that are distantly related within the induced topology.
Main Results:
- Topology-based selection demonstrated faster progress compared to dynamic subset selection and stochastic subset sampling.
- The method effectively utilizes problem structure information acquired during the evolutionary search.
Conclusions:
- The proposed topology-based subset selection method offers a significant improvement in the speed of genetic programming.
- This approach provides a robust and efficient way to manage large fitness case datasets in evolutionary algorithms.