Samsung AI Center – Montreal Researchers Win the Best Paper Award at IEEE GLOBECOM 2021

Researchers from Samsung AI Center – Montreal (SAIC-Montreal) received the Best Paper Award at the annual IEEE (Institute of Electrical and Electronics Engineers) Global Communications Conference (GLOBECOM) 2021 Mobile and Wireless Networks Symposium. GLOBECOM is the IEEE Communications Society’s flagship conference dedicated to driving innovation in nearly every aspect of communications. Each year, more than 3000 scientific researchers from academia and industry participate in a mutual exchange of new communication technologies.

The paper is resulted from the AI-based Methods for Intelligent Network project, involving stakeholders from Samsung’s Networking business unit (NBU), Samsung Research’s Advanced Communications Research Center, and SAIC-Montreal. SAIC-Montreal team is developing innovative AI technologies to improve the network performance of 5G telecommunications currently and of 6G in the near future. “We are honored to receive this best paper award from GLOBECOM. AI-based methods have proven to be exceptionally powerful in network management and operations. AI-based prediction algorithms we are developing are the perfect complement for our work on network management, and they help us better control and optimize the system’s performance,” said Dr. Steve Liu, VP R&D, Co-Director at SAIC- Montreal, who leads the project.

This award paper titled “One for All: Traffic Prediction at 5G Edge with Data-Efficient Transfer Learning” focuses on the prediction of 5G network traffic, which is one of the most important measures of 5G networks. By predicting the traffic volume at edge nodes (such as base stations), one can eliminate or reduce the network performance degradation caused by real-world system delays, and support many 5G edge operations in time, such as load balancing and automated network slicing. Existing prediction work cannot support such an edge-oriented framework, since resources and data on the distributed edge nodes are usually limited and imbalanced, which could bring serious overfitting and large prediction errors. To overcome this challenge, researchers from SAIC-Montreal proposed a Transfer Learning based Prediction framework that allows the edge nodes to share their resources and data efficiently, thus improving the prediction accuracy of data-limited edge nodes.

SAIC-Montreal is located in Montreal, the second largest city in Canada and home to one of the world’s fastest growing AI community. The center conducts research in machine learning, telecommunications, and robotics in tandem with the SRA Silicon Valley and Toronto AI Centers. “We are building an all-star team here in Montreal. I am incredibly proud of our team. We have been able to achieve great results in term of both academic recognition as well as practical value“, said Dr. Greg Dudek, VP R&D, Director of SAIC-Montreal.