On the Complexity of Chromatic Dispersion Compensation based on End-to-End Learning

Author(s)

Chaves, F. E. C. ; Rosa, E. S. ; Maciel, J. A. S. ; Sutili, T. ; Figueiredo, R. C.

• Document type: Journal article
• Source: 2024 SBFOTON INTERNATIONAL OPTICS AND PHOTONICS CONFERENCE, SBFOTON IOPC 2024; v. N/A, p. 3-pg., 2024-01-01.

Abstract

In order to verify and validate the use of machine learning techniques for chromatic dispersion compensation in optical transmissions, we elaborated an end-to-end recurrent neural network aiming at replacing the conventional digital signal processing (DSP) blocks employed at the transmitter and the receiver. Moreover, in order to establish a valid comparison between the conventional DSP and the integrated neural networks, we evaluate the computational complexity of each method for the same simulation scenario. Despite presenting a good performance for the compensation of chromatic dispersion, the end-to-end technique presented a significantly greater computational complexity, limiting its potential use in practical applications. (AU)

• FAPESP's process: 21/06569-1 - High-speed strategic internet technologies
• Grantee: Evandro Conforti
• Support Opportunities: Research Projects - Thematic Grants