A carregar...
Publicação
Towards a new geometric deflating crossover
| Nome: | Descrição: | Tamanho: | Formato: | |
|---|---|---|---|---|
| 553.84 KB | Adobe PDF |
Autores
Orientador(es)
Resumo(s)
Geometric Semantic Genetic Programming(GSGP) is a variant of Genetic Programming(GP) that utilizes geometric operators for supervised learning problems. These operators have the peculiarity of inducing a unimodal fitness landscape in the search space, removing local optima. The major drawback of this algorithm is that individuals grow rapidly through evolution, creating final models that are uninterpretable for humans, making GSGP a black box model. [1] Recently introduced by Vanneschi et al., Semantic Learning algorithm with Inflate and deflate Mutations (SLIM_GSGP), is a new variant of GSGP that employs a new geometric mutation, defined as deflate mutation. This new operator has the same geometric properties as the one used in GSGP, with the difference of creating new individuals that are smaller than their parent.[2, 3] This poster focuses on the proposal of new geometric crossovers, that exploit the new deflating principles introduces with SLIM_GSGP
Descrição
Farinati, D. (2024). Towards a new geometric deflating crossover: Utilizing the new linked-list structure of SLIM_GSGP individuals to develop new geometric operators [poster]. 1. Poster session presented at Data Research Meetup by MagIC, Lisbon, Portugal. --- This work was supported by national funds through FCT (Fundação para a Ciência e a Tecnologia), under the project - UIDB/04152/2020 (DOI: 10.54499/UIDB/04152/2020) - Centro de Investigação em Gestão de Informação (MagIC)/NOVA IMS).