Special Session: RobustFL

Towards Robust Federated Learning: Addressing Data and Device Heterogeneity

Organizers

Dr. Diletta Chiaro, University of Naples Federico II
Prof. Francesco Piccialli, University of Naples Federico II
Dr. Fabio Giampaolo, University of Naples Federico II

Primary Contact

Abstract

Federated Learning (FL) has emerged as a groundbreaking approach to enabling AI across decentralized and sensitive data sources without requiring data centralization, addressing privacy concerns that align with regulations like GDPR. For a leading conference like IJCNN, which emphasizes advancements in neural networks and transformative AI, FL offers a rich avenue for exploration. While past sessions at IJCNN (see IJCNN 2024) have largely focused on privacy and trustworthiness, the next critical frontier is addressing the inherent challenges of non-iid data distributions and device heterogeneity—barriers that must be overcome for FL to scale and perform effectively in real-world settings.

In practical FL deployments, data generated across decentralized sources is not only non-iid but is also produced by diverse devices with varying computational power, network stability, and memory capacity. This device heterogeneity complicates learning, as not all devices can contribute equally or reliably. Such variability can degrade model performance, create biases, and complicate both convergence and fairness. For example, FL applications in healthcare and IoT span a range of devices, from powerful servers to lightweight edge devices, each with unique data characteristics and resource constraints. Addressing non-iid data and device heterogeneity is essential for FL to be scalable, fair, and ethically viable, especially in critical areas like medical diagnostics and financial forecasting.

Important Dates

Submission Deadline: January 15, 2025

Scope and Topics of Interest

Algorithmic and Theoretical Advances: New algorithms that improve FL’s performance and adaptability to both non-iid data and heterogeneous device constraints.
Fairness and Robustness: Strategies for creating models that are fair and resilient across varying data sources and device capacities.
Interpretability and Explainability: Approaches to ensure transparency and interpretability in FL models that operate across diverse data and devices.
Personalization and Client Adaptivity: Techniques for personalizing FL models to specific clients while maintaining overall robustness, despite differing device capabilities and data.
Resource-Aware FL: Frameworks that optimize FL for resource-limited devices, accounting for constraints in memory, battery life, and connectivity.
Application Case Studies: Real-world applications demonstrating FL’s performance in managing non-iid data and device heterogeneity, particularly in sensitive areas like healthcare and IoT.
Tools and Benchmarks: Development of benchmark datasets and evaluation metrics that reflect the challenges of both non-iid data and diverse device conditions.

Program Committee

Afaf Taïk, Mila – Quebec AI Institute
Afsana Khan, Maastricht University
Alessio Mora, University of Bologna
Alessio Masano, Università degli studi di Catania
Bingjie Yan, Institute of Computing Technology, Chinese Academy of Sciences
Bostan Khan, Mälardalens University
Chaochao Sun, Shanghai University of Electric Power
Daniela Annunziata, University of Naples Federico II
Danilo Menegatti, Sapienza University of Rome
Depeng Chen, Anhui University
Diletta Chiaro, Università degli Studi di Napoli Federico II
Divya Kulkarni, Harvard Medical School – Massachusetts General Hospital
Enrique Tomás Martínez Beltrán, University of Murcia
Fabio Giampaolo, University of Naples Federico II
Giovanni Paragliola, CNR
Guanglei Yu, 新疆医科大学
Hanyue Xu, School of AI and Advanced Computing, Xi’an Jiaotong-Liverpool University
Haokun Chen, University of Munich
Jayant Vyas, Indian Institute of Technology Jodhpur
Jiayi Mao, Zhejiang University
Jing Xu, School of Computer and Communication Engineering, Changsha University of Science and Technology
Judith Sáinz-Pardo Díaz, Spanish National Research Council (CSIC)
Jue Xiao, Huazhong University of Science and Technology
Junfeng Chen, Jinan University
Kejia Zhang, School of Computer Science and Big Data (School of Cybersecurity) Heilongjiang University
Ljubomir Rokvic, EPFL
Manasa Mariam Mammen, Mercedes-Benz AG
Manuel Röder, Technical University of Applied Sciences Würzburg-Schweinfurt
Marco Garofalo, University of Pisa
Mario Colosi, University of Messina
Martina Savoia, University of Naples Federico II
Marzia Canzaniello, University of Naples Federico II
Melike Geçer, University of Lausanne
Mengchen Fan, University of the Alabama at Birmingham
Mingjia Shi, Sichuan University
Mingkai Hu, Henan University of Science and Technology
Monika Farsang, Technische Universität Wien (TU Wien)
Mrinmay Sen, Indian Institute of Technology Hyderabad
Nan Yang, The University of Sydney
Neena Goveas, BITS Pilani, Goa
Obaidullah Zaland, Umeå University
Oudom Kem, CEA
Paul Zheng, RWTH Aachen University
Pian Qi, University of Naples Federico II
Raman Zatsarenko, Rochester Institute of Technology
Renyou Xie, Central South University
Samuele Fonio, Università degli studi di Torino
Sheng Wan, The Hong Kong University of Science and Technology
Sileshi Nibret Zeleke, University of Bari, Department of Computer Science
Songlian Yan, 华南师范大学
Tianxiang Chen, Beihang University
Truong Thao Nguyen, National Institute of Advanced Industrial Science and Technology (AIST)
Wai Fong Tam, Queen Mary University of London
Xuwei Fan, Zhejiang Gongshang University
Yan Zhou, Kunsan National University
Yansong Zhao, Nanyang Technological University
Yichen Li, Huazhong University of Science and Technology
Yihang Wu, 桂林电子科技大学
Yiming Chen, Tsinghua University
Yitao Chen, Arizona State University
Youngjoon Lee, KAIST
Yu Huo, The Chinese University of Hong Kong, Shenzhen
Zahra Batool, Mila – Quebec AI Institute
Zhe Li, Rochester Institute of Technology