The Optical Networking and Communication
Conference & Exhibition

A Virtual Conference - Pacific Daylight Time (UTC-07:00)

The Role of Machine Learning in Optical Systems and the Role of Optics in Machine Learning Systems


S. J. Ben Yoo, University of California, Davis, USA; Manya Ghobadi, MIT, USA


This symposium explores optics, machine learning, and AI techniques for next-generation data centers and networking systems, both in terms of leveraging ML techniques for optical systems (ML for optics) and optical technologies for ML workloads (optics for ML) and their impacts, and use cases. In this Symposium, large-scale network operators, ML/AI experts, virtualization researchers, and technology researchers will discuss (a) the role of machine learning and AI techniques to effectively control, manage, and plan next-generation data centers, high-performance clusters, and communication networks, and (b) the role of emerging optical technologies in enabling disaggregated ML systems, next-generation ML/AI architectures, application-driven reconfiguration,  neuromorphic computing. The Symposium will include a panel discussion on emerging photonic and photonic-electronic systems for efficiently and effectively accelerating ML training and inference workloads.  


Part 1:  Machine Learning and AI in Data Centers and Optical Networks

Katharine Schmidtke, Facebook, USA
Applied Machine Learning in Facebook Data Centers

Norm Jouppi, Google, USA
Machine Learning Workload and TPU in Data Centers

Hitesh Balani, Microsoft, UK
The Role of Photonic Switching for Machine Learning Workloads in Data Centers

Roberto Proietti, University of California, Davis, USA
Machine-Learning-Aided Bandwidth and Topology Reconfiguration for Optical Data Center Networks

Panel discussion 

Part 2: Photonics in Machine Learning and AI for Future Data Centers

Larry Dennison, NVIDIA, USA
What Photonics can do in GPU Based Machine Learning Systems

Dirk Englund, MIT, USA
Photonic Neural Networks

Bert Offrein, IBM, Switzerland
Photonic Integrated Circuits for Neural Network Inference and Training

Chris Cole; II-VI Incorporated, USA
Optical and Electrical Computing Energy Use Comparison

Panel discussion 

Sponsored by: