Machine Learning For Optical Communication Systems

NIST will hold a workshop at the Boulder Colorado Laboratories to discuss the role of machine learning (ML) in optical communication systems. Optical communication systems are increasingly used closer to the network edge and are expected to find use in new applications that require more intelligent functionality. Optical networks are needed to address the high speeds and low latency of 5G wireless networks. The analog nature of optical transmission and the complexity of operation and management remain an impediment to greater use of software controls. Furthermore, optical systems are running up against spectral density limits that threaten traditional capacity-based scaling. New efficiency-based scaling methods are needed to further improve the cost/bit/s without relying on capacity improvements alone. Artificial intelligence (AI) and machine learning provide a new direction with the potential to both enable wider use of software controls and to further optimize the efficiency of optical systems across multiple dimensions. Reference data sets for ML would improve functionality and operability across industry further enabling scaling and efficiency.  This workshop seeks to identify and develop applications of AI, and ML in the context of accelerating the use of software-based networking in optical systems for improved performance and scalability. Paths to realizing reference training data sets for ML in optical communications systems including needs for new or different metrology will be examined. Workshop outcomes include a white paper for a plan and path to develop and disseminate reference data sets for ML training and applications to open and programmable optical communication systems, as well as a working group to further develop these ideas.

Update

Just released on March 27th , 2020 for public dissemination a NIST Special Publication Series 2100 White Paper. The publication titled: “Summary: Workshop on Machine Learning for Optical Communication Systems” provides a summary and outcomes from the NIST workshop on Machine Learning For Optical Communication Systems that was hosted on August 2nd, 2019 by the NIST Communications Technology Laboratory (CTL) hosted in partnership with the Laboratory for Information Technologies (ITL). This workshop brought together industry, academia, and government to discuss the roll of artificial intelligence and machine learning in optical communication systems. This NIST special Publication White Paper provides information reflecting the presentations and break-out session discussions which took place during the workshop. The publication also provides a consolidation of important themes related to the topic of machine learning and its application to optical communications systems to address important use cases such as 5G communications, network management, resiliency, scalability and requirements for data sets for machine learning algorithms in the context of optical communications systems. References relevant to information contained in the Special Publication are also provided.


Presentation Slides from August 2, 2019 Workshop on Machine Learning for Optical Communications Systems

• Machine Learning Applications in Optical Transport Networks
• ROADM and Optical Layer
• Cross-layer/Multi-vendor end-to-end networking
• Small Data Deep Learning: AI Applied to Domain Datasets
• What data matters for optical communications
• Machine Learning Datasets from Optical Systems
• Machine Learning Models for Optical Communication Systems and Networks