What is the Role of Machine Learning in Optical Access Networks


Michael Freiberger, Verizon Communications Inc., USA
Marco Ruffini, Univ. of Dublin Trinity College, Ireland
Junwen Zhang, Fudan Univ., China


Artificial Intelligence (AI) and Machine Learning (ML) have attracted increasing interests across a wide range of applications for optical access networks. From physical-layer transmission to wavelength routing, we have experienced over the past decade an increase in complexity of optical access networks, due to increasing data transmission speed, more dynamic and connections and more complicated use cases.

AI and ML have shown promising results for optimization, prediction and identification in systems that exhibit nonlinear, dynamic and complex behaviors and are thus good candidates to tackle optical access networks problems. For instance, recently studies have shown ML algorithms can improve the transmission performance by non-linearity impairments compensation. It has also been reported that ML can achieve better efficiency for bandwidth allocations. We have also seen promising results of using ML for pro-active virtual topology management, efficient network operations, and scalable network automation in access networks.

This panel will provide a forum for a wide range of speakers to share their ideas on ML/AI over novel applications in optical access networks. The panel will discuss the use of AI and ML in areas such as: DSP for signal impairment compensation; dynamic network capacity allocation; network management and resilience; optical/wireless and access/metro integration; in-home intelligent access network and more. 


To be determined.