DC9 Project: Decentralised Graph Neural Networks: Adaptation, Training and Interpretability in Federated Environments

Doctoral Candidate

S M Asiful Huda, MSc

Main Supervisor: Salvatore Cuomo (UNINA)

Co-Supervisors: David Camacho (UPM), Shen Yin (NTNU), Dariusz Mrozek (SUT)

R&D cooperation: ALMAWAVE

Objectives: advancing the field of Graph Neural Networks (GNNs) within federated and decentralised learning frameworks. 

Graph-based data structures are increasingly relevant for representing complex relationships in domains such as transportation, communication networks, and social systems. However, their distributed and interconnected nature poses unique challenges for privacy, scalability, and interpretability. The main goal of this research is to design and implement federated GNN architectures capable of efficient training across decentralised nodes, without requiring central data aggregation. The doctoral candidate will investigate graph partitioning strategies, federated aggregation mechanisms, and model interpretability methods specifically tailored for structured, non-IID data. A major focus will be on explainability, ensuring that the decision processes of decentralised GNNs are transparent and comprehensible to human users and domain experts.

The research will combine theoretical innovation with applied validation through collaboration with industrial partners and academic institutions. Expected accomplishments include the design of a prototype federated GNN framework, a set of interpretability tools for decentralised AI systems, and comprehensive guidelines for deploying GNNs in real-world distributed environments.

Expected Results: 

Development of a federated GNN prototype capable of learning over decentralised graph-structured datasets.

Establishment of methods for interpretability and explainability of distributed GNN decisions.

Benchmarking of federated GNNs against centralised models to evaluate performance, efficiency, and privacy.

Publication of high-impact papers in top-tier AI conferences and journals.

Applied contributions to the TUAI network’s industrial partner ALMAWAVE S.p.A., focusing on scalable, explainable AI systems for complex data analytics.

Applied research: The applied research in this project will develop and refine a Graph Neural Network (GNN) prototype suited for federated learning environments, focusing on handling decentralized graph-structured data. The project aims to innovate in training GNNs without centralizing data, ensuring efficient data processing with minimal information leakage. A significant focus will be on enhancing GNN interpretability in a decentralized context, making these models transparent and easily understandable. The project will culminate in a practical GNN framework, offering insights and best practices for GNN adaptation in federated settings, bridging theoretical research with practical applications in fields requiring complex data analysis.

Planned secondments: UPM(4 months); SUT(4 months); NTNU (4 months)                   

Enrolment in Doctoral degree: UNINA