Related Experiment Videos
A controlled genetic programming approach for the deceptive domain
Emin Erkan Korkmaz1, Göktürk Uçoluk
1Department of Computer Engineering, Middle East Technical University, 06531 Ankara, Turkey. korkmaz@ceng.metu.edu.tr
Summary
This study introduces a new method to guide genetic programming (GP) recombination by using global information from promising solutions. This approach aims to improve search efficiency and speed up convergence in genetic algorithms.
Area of Science:
- Computer Science
- Artificial Intelligence
- Evolutionary Computation
Background:
- Traditional genetic programming (GP) relies on random subtree combination via crossover, often leading to slow convergence.
- There is a need for controlled recombination strategies to enhance the efficiency of GP algorithms.
Purpose of the Study:
- To present a novel approach for guiding the recombination process in genetic programming.
- To leverage global information from promising solutions to control crossover operations.
- To improve the convergence speed of genetic search.
Main Methods:
- Extracting global information from high-performing solutions identified during the genetic search.
- Developing a separate control module to process this extracted information.
- Implementing a feedback mechanism from the control module to the genetic engine to guide recombination choices.
Main Results:
- The proposed method provides a controlled way to guide the crossover operation in GP.
- The control module analyzes solution information to inform recombination decisions.
- Feedback on potential recombination consequences is provided to the genetic engine.
Conclusions:
- The presented approach offers a method to enhance genetic programming by intelligently guiding recombination.
- Utilizing global solution information can lead to more efficient and faster genetic searches.
- This controlled recombination strategy holds promise for improving GP performance.