RISE Research Radar

Computer Science Open House 2022-2025

2025

Generalizable Representation for Wireless Networks through Native Graph Topology

Yifei Jin, Sarunas Girdzijauskas, Aristides Gionis

Summary

Uses graph representation learning and transformer architectures for wireless network optimization in 5G/6G. Award-winning solution in ITU AI/ML in 5G Challenge with 140+ competitors. Achieves real-time traffic congestion and E2E latency estimation, radio coverage optimization faster than commercial simulators, and wireless ray-tracing surrogates.

Themes

wireless-networksmachine-learning

Keywords

graph neural networks, 5G, 6G, traffic prediction, ray tracing, network optimization

Poster

Generalizable Representation for Wireless Networks through Native Graph Topology poster

Click image to open full size