Revisiting Multi-Step Nonlinearity Compensation with Machine Learning

• DOI: 10.48550/arxiv.1904.09807
• Publication Date: 2019-01-01
• Authors: Christian Häger; Henry D. Pfister; Rick M. Bütler; Gabriele Liga; Alex Alvarado
• Keywords: Machine Learning Based Dsp; Signal Processing (eess.SP); FOS: Computer and information sciences; Low-Complexity Digital Back-Propagation; Computer Science - Artificial Intelligence; Computer Science - Information Theory; Information Theory (cs.IT); Communication Systems; 0211 other engineering and technologies; Machine Learning (stat.ML); 02 engineering and technology; Subband Processing; Polarization Mode Dispersion; Deep Learning; Artificial Intelligence (cs.AI); Statistics - Machine Learning; Signal Processing; Telecommunications; FOS: Electrical engineering, electronic engineering, information engineering; 0202 electrical engineering, electronic engineering, information engineering; Electrical Engineering and Systems Science - Signal Processing

For the efficient compensation of fiber nonlinearity, one guiding principles appears to be: fewer steps are better and more efficient. We challenge this assumption show that carefully designed multi-step approaches can lead performance-complexity trade-offs than their few-step counterparts.