Título: A Hybrid Approach of Grammar-based Genetic Programming and Differential Evolution for Symbolic Regression
Autores: Andrade Motta, Flávio;Bernardino, Heder;Barbosa, Helio;de Freitas, João;Oliveira, Itamar;Souza, Felipe
Genetic Programming (GP) is used for solving many real world problems. From data classification to building phylogenetic trees, the technique can be applied almost to any problem. One way to improve GP performance is using a formal grammar. We propose here the use of grammar-based genetic programming (GGP) with Differential Evolution (DE). DE is incorporated to GGP in order to improve the quality of solutions obtained by GGP when solving symbolic regression problems by finding good numerical coefficients for the models. In this proposal, the coefficients of the best individuals generated by GGP during the search are adjusted by DE. Also, this technique incorporates these values to the grammar; thus, the grammar is adapted during the search. The proposed technique is applied to 8 symbolic regression problems and it is compared to a standard GGP. The results indicate that GGP hybridized with DE obtained better models, specially when the original model contains real-valued coefficients.
Grammar-based Genetic Programming;Differential Evolution;Hybridism;Symbolic Regression
Código DOI: 10.21528/CBIC2017-110
Artigo em pdf: cbic-paper-110.pdf
Arquivo BibTeX: cbic-paper-110.bib