Geometric Deep Learning: from grid to graph structured data

You are cordially invited to attend the PhD Dissertation Defence of Federico Monti on Thursday 23 January 2025 at 10:00 in room D0.03.

Abstract:
The success of Deep Learning architectures (e.g. Convolutional Neural Networks, Recurrent Neural Networks, Transformer Networks, . . .) and the increasing availability of graph / manifold structured data (e.g. social networks, sensor networks, molecules, 3D shapes, . . .) motivated, in recent years, the development of a new class of Geometric Deep Learning (GDL) approaches aimed at extending traditional DL solutions to non-Euclidean domains. In this talk, we will explore the realm of Graph Convolutional Neural Networks (GCNNs), a popular class of GDL architectures that rely on generalizations of the convolution operation to process the provided input data. In the first part of the presentation, we will focus on methodologies. We will review the two fundamental classes of approaches (spatial and spectral constructions) that can be used for generalizing convolution on graphs, we will introduce two different GCNNs that respectively rely on an attention mechanism (MoNet) and complex rational spectral filters (CayleyNet) to implement graph convolutional layers, and we will highlight the performance that the proposed methods allow to achieve on shape correspondence and document classification tasks. In the second part of the presentation, we will focus instead on applications of GCNNs, and we will discuss in particular how this specific class of approaches can be used to effectively detect misinformation based on the patterns that article / tweets form when spreading on Twitter / X. The content of the second part of the talk corresponds with the core technology of Fabula AI (our own startup that was acquired by Twitter in 2019), and it represents an example of how GCNNs can be successfully applied in an industrial setting.

Dissertation Committee:
- Prof. Michael Bronstein, University of Oxford, UK and AITHYRA, Austria (Research Advisor)
- Prof. Cesare Alippi, Università della Svizzera italiana, Switzerland (Internal Member)
- Prof. Andrea Emilio Rizzoli, Università della Svizzera italiana, Switzerland (Internal Member)
- Prof. Xavier Bresson, NUS, Singapore (External Member)