Andrea CARENA  (Politecnico di Torino)

ML models for the abstraction of photonic components in UWB systems

Abstract
In recent years data traffic has seen an extraordinary growth driven by a continuous increase in the number of connected devices and the development of new bandwidth hungry applications. To improve the capacity of optical networks, a promising solution is the adoption of Ultra-Wide-Band (UWB) systems expanding the fiber bandwidth beyond the standard C-band. In this context, the introduction of a software-defined networking (SDN) paradigm is a viable solution to deliver flexibility and dynamic reconfigurability to the network. To implement SDN in UWB networks it is required the full abstraction and virtualization of each network element as it allows operations coordinated by a centralized network controller. This objective can be reached by defining simple but accurate models for all components. This talk presents an approach based on the application of Machine Learning (ML) techniques and it focus on two network elements: the photonic switch and the Raman amplifier.