Comparison of Different Computing Platforms for Implementing Parallel Genetic Programming

Ruihua Zeng, Zhixing Huang, Yongliang Chen, Jinghui Zhong, Liang Feng

Abstract: Genetic programming (GP) is a powerful tool for knowledge discovery and data mining. Over the past decades, GP has been implemented in various parallel computing platforms to reduce its search time. However, these parallel GPs have different design principles and performance characteristics, which makes it difficult for users to choose the proper parallel GP in practice. To address this issue, this paper focuses on comparing and analyzing the characteristics of parallel GPs implemented in different computing platforms, in terms of running time, the speedup ratio, and the scalability. Based on the empirical results, the guidance of selecting different parallel GPs is concluded.