Anzeige
Erschienen in:
2021 | OriginalPaper | Buchkapitel
Compressive Autoencoders in Wireless Communications
verfasst von : Peijun Chen, Peng Lv, Hongfu Liu, Bin Li, Chenglin Zhao, Xiang Wang
Erschienen in: Communications, Signal Processing, and Systems
Verlag: Springer Singapore
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
Autoencoder, one of the most promising and successful architectures in deep learning (DL), has been widely used in wireless communications. However, fast-increasing size of neural networks (NNs) in autoencoder leads to high storage requirement and heavy computational overhead, which poses a challenge to practical deployment of autoencoders in real communications systems. In this paper, we investigate two representative NNs compression methods and propose two compressive autoencoder schemes for wireless communications by combining the compression techniques with autoencoder architecture. Our proposed schemes are capable to reduce memory consumption and execution time. Numerical experiments demonstrate that our methods can effectively compress the autoencoder’s size without degrading the model performance or distorting the constellations of transmitted signal.