[TuP-A] Poster Session 1

[TuP-A-5] Two-Stage Federated Learning-Based Aggregated Traffic Prediction in Edge Computing-Enabled Metro Optical Networks

Citong Que¹, ○Hongcheng Wu¹, Faisal Nadeem Khan¹ (1. Tsinghua University (China))

Keywords:

Artificial intelligence and machine learning for optical network design,control,and management

We propose a hierarchical model aggregation-based federated learning framework that handles heterogeneous and imbalanced data for traffic prediction in edge computing-enabled metro optical networks (MONs). It reduces the prediction errors by ~5% for large-scale MONs.