Soft target and functional complexity reduction: A hybrid regularization method for genetic programming
作者：
Highlights：
• Definition of soft target regularization for genetic programming (GP).
• Definition of a novel measure of functional complexity for GP individuals.
• Reduction of overfitting by using the soft regularization and complexity measure.
• Definition of a hybrid system that removes overfitting in all the benchmarks.
• Study of the relation between generalization, complexity, and individuals’ size.
摘要
•Definition of soft target regularization for genetic programming (GP).•Definition of a novel measure of functional complexity for GP individuals.•Reduction of overfitting by using the soft regularization and complexity measure.•Definition of a hybrid system that removes overfitting in all the benchmarks.•Study of the relation between generalization, complexity, and individuals’ size.
论文关键词：Regularization,Genetic programming,Soft target,Functional complexity,Hybrid system
论文评审过程：Received 15 June 2020, Revised 19 January 2021, Accepted 19 March 2021, Available online 27 March 2021, Version of Record 13 April 2021.
论文官网地址：https://doi.org/10.1016/j.eswa.2021.114929
