BEGIN:VCALENDAR
PRODID:-//eluceo/ical//2.0/EN
VERSION:2.0
CALSCALE:GREGORIAN
BEGIN:VEVENT
UID:43c41100c4822bcd86d664dcf6067d7c
DTSTAMP:20261024T012616Z
SUMMARY:I-X Seminar Series: Understanding message passing: limitations of t
 he paradigm and new possibilities with Francesco Di Giovanni
DESCRIPTION:Lecture Title\nUnderstanding message passing: limitations of th
 e paradigm and new possibilities\nSpeaker\nFrancesco Di Giovanni\nTalk Sum
 mary\nMessage passing (MP) stands as a cornerstone in Geometric Deep Learn
 ing\, driving the success of Graph Neural Networks (GNNs) in analysing bot
 h graphs and point clouds\, and leading to empirical achievements in many 
 scientific domains. Despite its widespread application\, the theoretical u
 nderpinnings of MP’s successes and limitations remain underexplored. Thi
 s talk delves into a key systematic limitation known as over-squashing\, w
 hich significantly impacts problems involving long-range interactions. I w
 ill discuss how over-squashing emerges from a disconnect between data topo
 logy and the intended task\, highlighting its role in revealing shortcomin
 gs of existing approaches to study the expressive power of GNN models. A m
 ajor focus will be on graph rewiring\, a novel approach in MP that treats 
 topology as a critical design element\, offering a solution to mitigate ov
 er-squashing and expand the capabilities of MP. The talk will conclude by 
 exploring exciting future directions\, including applications in generativ
 e AI and the sparsification of large Transformer models\, promising new ho
 rizons for MP’s evolution in Geometric Deep Learning.\nSpeaker Bio\nI am
  a Senior Research Associate in the Department of Computer Science at the 
 University of Oxford\, working under the supervision of Michael Bronstein.
  Specialising in Geometric Deep Learning (GDL)\, my research has primarily
  revolved around Graph Neural Networks. I have made contributions in ident
 ifying limitations of the current message-passing paradigm and have develo
 ped extensions to enhance their efficacy\, particularly for tasks involvin
 g long-range interactions. My current research also spans generative AI an
 d a theoretical reassessment of conventional GDL methodologies. Before joi
 ning Oxford\, I served as a Research Associate at the University of Cambri
 dge and as a Machine Learning Researcher at Twitter. I earned my PhD in Ma
 thematics from UCL in 2021\, under the guidance of Jason Lotay\, with a th
 esis on Ricci flow. My general goal is understanding how symmetries and ge
 ometry of the data affect the power and optimization of deep learning mode
 ls\, with the ambition of working towards the next generation of AI for no
 n-Euclidean domains\, which are common in Life Sciences.\nTime: 13.00 – 
 14.00\nDate: Thursday 8 February\nLocation: In person | I-X Conference Roo
 m | Level 5 | Translation and Innovation Hub (I-HUB)\nImperial White City 
 Campus\n84 Wood Lane\nW12 0BZ
URL:https://www.imperial.ac.uk/events/173285/i-x-seminar-series-understandi
 ng-message-passing-limitations-of-the-paradigm-and-new-possibilities-with-
 francesco-di-giovanni/
DTSTART;TZID=Europe/London:20240208T130000
DTEND;TZID=Europe/London:20240208T140000
LOCATION:Level 5 | I-X Conference Room\, Translation and Innovation Hub (I-
 HUB)\, White City Campus\, Imperial College London\, London\, W12 0BZ\, Un
 ited Kingdom
END:VEVENT
BEGIN:VTIMEZONE
TZID:Europe/London
BEGIN:STANDARD
DTSTART:20240208T130000
TZNAME:GMT
TZOFFSETTO:+0000
TZOFFSETFROM:+0000
END:STANDARD
END:VTIMEZONE
END:VCALENDAR