Anzeige
Erschienen in:
2020 | OriginalPaper | Buchkapitel
An Evolutionary Approach to Hyper-Parameter Optimization of Neural Networks
verfasst von : Marco Stang, Christopher Meier, Vinzenz Rau, Eric Sax
Erschienen in: Human Interaction and Emerging Technologies
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
Abstract
The last few years have increasingly been influenced by the emergence of neural networks for classification tasks. Due to the increasing size of the nets, especially concerning deep neural nets, the selection of the optimal topology becomes critical. A set of hyper-parameters determines the topology of a neural network. The selection of suitable hyper-parameters however, is time-consuming or requires expert knowledge. This paper examines the application of an evolutionary algorithm to optimize the hyper-parameters of a neural network. The result of the optimization by evolutionary algorithms can be applied for a variety of use-cases: i.e. detection of anomalies in CAN-bus data, object recognition for autonomous driving or the classification of hydraulic machinery oil.