(c) depositphotos.com

9th International Conference on Machine Learning for Networking (MLN'2026)


Paris, France, December 1-3, 2026

Keynote: Anastasios Giovanidis, Ericsson Research & CNRS, France


Title: Foundation and World Models for Autonomous Telecom Networks


Abstract

The telecom industry is moving from reactive network management toward autonomous operations. AI is central to this transition, yet today's ML models remain narrow — each scoped to a single KPI, domain, or deployment. Foundation models, pre-trained on diverse data and adaptable to many downstream tasks, offer a path beyond this fragmentation.

Building such models for telecom is challenging. Network data is structured and tabular, with hundreds of configuration and performance counters per cell. It is relational — cells form a spatial topology with heterogeneous neighbor relations reflecting interference, load sharing, and coverage. And it is temporal, with metrics evolving at sub-hourly granularity across daily and weekly cycles. A telecom foundation model must learn jointly from all three aspects.

Yet prediction alone does not enable autonomy. World models predict how network state evolves in response to candidate actions, enabling what-if scenario analysis, intent-based management, and digital twins. When built on foundation model representations, these become foundation world models — a general-purpose engine for autonomous decision-making.

This talk surveys the emerging landscape and discusses open challenges including data availability, action-conditional training, spatial generalization, and efficient deployment. We argue that foundation world models are a key enabler for network autonomy, and that collaboration between vendors and service providers is essential to realize this vision.


Biography

Anastasios Giovanidis is a Master Researcher on AI at the Ericsson R&D center in Massy, France, working on Reinforcement Learning methods and their application to performance analysis and optimization of cellular networks. He is also a tenured researcher of the French National Centre for Scientific Research (CNRS), on-leave since 2023 and a visiting researcher at Sorbonne University — LIP6 lab. As CNRS member, Anastasios was affiliated with the LIP6 Laboratory of Sorbonne University, and the LTCI laboratory of Telecom ParisTech. He has held research positions at the Fraunhofer Heinrich Hertz Institute in Berlin, the Zuse Institute Berlin and the INRIA institute in Paris. He received the Diploma in Electrical and Computer Engineering from the National Technical University of Athens, Greece, and the Dr.-Ing. degree from the Technical University of Berlin, Germany. Currently, his research focuses on Machine Learning, AI, Reinforcement Learning, Graph Neural Networks, 6G Telecom Networks L2 & L3, Energy efficiency, Data Mining and Probabilistic Models.