Training strategies to minimize interchannel interference effects using supervised learning in gridless Nyquist-WDM...

Area of Science:

• Optical communication networks
• Machine learning applications
• Signal processing

Background:

• Future optical networks demand advanced distortion mitigation techniques.
• Interchannel interference (ICI) is a significant challenge in gridless WDM systems.
• Machine learning (ML) shows promise for addressing optical impairments.

Purpose of the Study:

• To propose and evaluate two ML training strategies for minimizing ICI in a 16-QAM Nyquist-WDM system.
• To assess the effectiveness of different ML algorithms (ANN, SVM, KNN, ELM) under these strategies.
• To analyze the performance gains and computational complexity of ML-based ICI mitigation.

Main Methods:

• Investigated two supervised learning training strategies: updating (symbol-based) and characterization (offline).
• Applied Artificial Neural Networks (ANN), Support Vector Machine (SVM), K-Nearest Neighbors (KNN), and Extreme Learning Machine (ELM).
• Evaluated performance using Bit Error Rate (BER) and Optical Signal-to-Noise Ratio (OSNR) in back-to-back and 250 km fiber transmission scenarios.

Main Results:

• Both training strategies improved BER, with up to 4 dB OSNR gains in back-to-back tests.
• KNN and ELM algorithms demonstrated significant BER reduction over 250 km of optical fiber.
• ELM exhibited low computational complexity, requiring only 1.9% of ANN processing time.

Conclusions:

• ML-based ICI mitigation techniques enhance performance in gridless optical networks.
• The proposed training strategies and algorithms are effective for reducing optical impairments.
• These advancements support the development of high-capacity networks to meet future traffic demands.