Machine Learning for Future Fiber-Optic Communication Systems

Chapter Five: End-to-end learning for fiber-optic communication systems

Ognjen Jovanovic; Francesco Da Ros; Metodi Yankov; Darko Zibar    Department of Photonic Engineering, Technical University of Denmark, Kgs. Lyngby, Denmark

Abstract

In this chapter, we review the application of end-to-end learning in optical communication systems. First, we briefly discuss the motivation and idea behind end-to-end learning using autoencoders. Later on, we illustrate applicability of end-to-end learning for geometric constellation shaping when applied to the approximate perturbation channel models. Furthermore, we discuss waveform optimization for dispersive channel focusing only on intensity modulation. We discuss end-to-end learning of waveforms when communicating ...