Talk: Javad Aliakbari (October 27, 2025 at 2:00 PM, Seminar room N2409)

From Eigenvalues to Learning: The Power of Spectral Graph Theory in Modern Machine Learning

Javad Aliakbari

Chalmers University of Technology

 

Abstract:

Spectral graph theory provides a powerful mathematical framework for understanding the structure and dynamics of complex networks through the eigenvalues and eigenvectors of graph matrices such as the Laplacian or adjacency matrix. In this talk, I will introduce the fundamental principles of spectral graph theory and demonstrate how they form the foundation of several key techniques in modern data analysis and machine learning. We will explore how spectral embeddings reveal hidden community structures in graphs (as in spectral clustering) and how these concepts extend naturally to graph neural networks (GNNs), where spectral filters enable learning directly in the frequency domain of graphs. Finally, I will discuss recent research directions that connect spectral methods with graph coarsening, graph signal processing, and federated graph learning, highlighting their implications for scalable and interpretable learning on large, distributed graphs.

Biography:

Javad Aliakbari is a PhD student at the Department of Electrical Engineering, Chalmers University of Technology, under the supervision of Prof. Alexandre Graell i Amat. His research focuses on federated and distributed graph learning, spectral methods, and privacy-preserving machine learning. He has co-authored several works on subgraph federated learning and spectral graph analysis, with applications in communication systems and large-scale networked data.