Machine learning for quality of transmission: a picture of the benefits fairness when planning WDM networks

• DOI: 10.1364/jocn.433412
• Publication Date: 2021-10-05T16:15:15Z
• Authors: Matteo Lonardi; Jelena Pesic; Thierry Zami; Emmanuel Seve; Nicola Rossi
• Keywords: 0202 electrical engineering, electronic engineering, information engineering; 02 engineering and technology

In the present day, evaluation of machine learning (ML) as a candidate for substituting analytical quality transmission (QoT) estimators is done in compartmentalized way. The assessment not produced from global optical network design perspective and with accurate metrics; on contrary, heavily focuses physical layer impairment precision capabilities while underemphasizing effects at layer. this paper, we recommend suitable methodology evaluating QoT substitution based foundational idea that different should be examined comparative basis by analyzing network-relevant metrics parity availability performance. Pragmatically, comparing performance estimation solutions through aggregate throughput, i.e., capacity, equity their overestimation likelihood, which drives system margins. To demonstrate need such viewpoint illustrate potential drawbacks an inadequate substitution, use proposed method several scenarios (altering topologies, input parameter uncertainty conditions, requirements), showing can achieve gains error or margins observing notable losses terms throughput. Considering results were contrary to what one may expect, decided undergo statistical analysis order investigate grasp consequences model distribution relation capacity.