Improving gene expression programming performance by using differential evolution

Gene Expression Programming (GEP) is an evolutionary algorithm that incorporates both the idea of a simple, linear chromosome of fixed length used in Genetic Algorithms (GAs) and the tree structure of different sizes and shapes used in Genetic Programming (GP). As with other GP algorithms, GEP has difficulty finding appropriate numeric constants for terminal nodes in the expression trees. In this work, we describe a new approach of constant generation using Differential Evolution (DE), a real-valued GA robust and efficient at parameter optimization. Our experimental results on two symbolic regression problems show that the approach significantly improves the performance of the GEP algorithm. The proposed approach can be easily extended to other Genetic Programming variations.

• Type: Conference
• Year: 2007
• Where: ICMLA
• Authors: Qiongyun Zhang, Chi Zhou, Weimin Xiao, Peter C. Nelson
• Keywords: Gene Expression Programming, Genetic Programming, GEP Algorithm, ICMLA 2007, Machine Learning