2011 | OriginalPaper | Buchkapitel
MetaHybrid: Combining Metamodels and Gradient-Based Techniques in a Hybrid Multi-Objective Genetic Algorithm
verfasst von : Alessandro Turco
Erschienen in: Learning and Intelligent Optimization
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
We propose a metamodel approach to the approximation of functions gradients within a hybrid genetic algorithm. The underlying structure is implemented in order to support parallel execution of the code: a genetic and a SQP algorithm run in different threads and can ask designs evaluations independently, but keeping all the available resources always working. A common archive collects the results and generates the population for the GA and the starting points for the SQP runs. A particular attention is dedicated to elitism and to constraints. The hybridization is performed through a modified
ε
−constrained method. The general philosophy of the algorithm is to concentrate on not wasting information: metamodels, archiving and elitism, steady-state parallel evolution are key elements for this scope and they will be discussed in details. A preliminary but explanatory row of tests concludes the paper highlighting the benefits of this new approach.