Abstract: Genetic programming is a powerful evolutionary algorithm that solves user-defined task by the evolution of computer programs. In the literature, GP has successfully been applied to improve deep learning (e.g., to optimize the structure of deep neural networks). However, little work has been reported on applying deep learning to improve EA. In this project, we attempt to utilize deep learning to improve GP. The key idea is to utilize deep learning techniques to identify important primitives (i.e., terminals and functions), so as to reduce the search space and to improve the search efficiency of GP.

Related paper:

  1. J. Zhong*, Y. Lin, C. Lu, and Z. Huang, “A Deep Learning Assisted Gene Expression Programming Framework for Symbolic Regression Problems”, ICONIP(7) 2018:530-541 . [Link]