2013 | OriginalPaper | Buchkapitel
On GPU Based Fitness Evaluation with Decoupled Training Partition Cardinality
verfasst von : Jazz Alyxzander Turner-Baggs, Malcolm I. Heywood
Erschienen in: Applications of Evolutionary Computation
Verlag: Springer Berlin Heidelberg
Aktivieren Sie unsere intelligente Suche, um passende Fachinhalte oder Patente zu finden.
Wählen Sie Textabschnitte aus um mit Künstlicher Intelligenz passenden Patente zu finden. powered by
Markieren Sie Textabschnitte, um KI-gestützt weitere passende Inhalte zu finden. powered by
GPU acceleration of increasingly complex variants of evolutionary frameworks typically assume that all the training data used during evolution resides on the GPU. Such an assumption places limits on the style of application to which evolutionary computation can be applied. Conversely, several coevolutionary frameworks explicitly decouple fitness evaluation from the size of the training partition. Thus, a subset of training exemplars is coevolved with the population of evolved individuals. In this work we articulate the design decisions necessary to support Pareto archiving for Genetic Programming under a commodity GPU platform. Benchmarking of corresponding CPU and GPU implementations demonstrates that the GPU platform is still capable of providing a times ten reduction in computation time.