Structural Risk Minimisation-Driven Genetic Programming for Enhancing Generalisation in Symbolic Regression

Created by W.Langdon from gp-bibliography.bib Revision:1.7615

@Article{Chen:ieeeTEVC,

• author = "Qi Chen and Mengjie Zhang and Bing Xue",
• journal = "IEEE Transactions on Evolutionary Computation",
• title = "Structural Risk Minimisation-Driven Genetic Programming for Enhancing Generalisation in Symbolic Regression",
• year = "2019",
• volume = "23",
• number = "4",
• pages = "703--717",
• month = aug,
• keywords = "genetic algorithms, genetic programming, Symbolic Regression, Generalisation, Structural Risk Minimisation, Vapnik-Chervonenkis Dimension",
• DOI = "doi:10.1109/TEVC.2018.2881392",
• ISSN = "1089-778X",
• size = "15 pages",
• abstract = "Generalisation ability, which reflects the prediction ability of a learnt model, is an important property in genetic programming for symbolic regression. Structural risk minimisation is a framework providing a reliable estimation of the generalisation performance of prediction models. Introducing the framework into genetic programming has the potential to drive the evolutionary process towards models with good generalisation performance. However, this is tough due to the difficulty in obtaining the Vapnik-Chervonenkis dimension of nonlinear models. To address this difficulty, this paper proposes a structural risk minimisation-driven genetic programming approach, which uses an experimental method (instead of theoretical estimation) to measure the Vapnik-Chervonenkis dimension of a mixture of linear and nonlinear regression models for the first time. The experimental method has been conducted using uniform and non-uniform settings. The results show that our method has impressive generalisation gains over standard genetic programming and genetic programming with the 0.632 bootstrap, and that the proposed method using the non-uniform setting has further improvement than its counterpart using the uniform setting. Further analyses reveal that the proposed method can evolve more compact models, and that the behavioural difference between these compact models and the target models is much smaller than their counterparts evolved by the other genetic programming methods.",
• notes = "Also known as \cite{8536418}",

}

Genetic Programming entries for Qi Chen Mengjie Zhang Bing Xue

Citations