BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Memento EPFL//
BEGIN:VEVENT
SUMMARY:Generalizing Graph Diffusion Models
DTSTART:20240202T130000
DTEND:20240202T150000
DTSTAMP:20260407T010130Z
UID:ddc980cbe421f7de38a2cd9d5ae51d293a25f4e828cbc5e1a84cd668
CATEGORIES:Conferences - Seminars
DESCRIPTION:Yiming Qin\nEDIC candidacy exam\nExam president: Prof. Maria B
 rbic\nThesis advisor: Prof. Pascal Frossard\nCo-examiner: Prof. Olga Fink\
 n\nAbstract\nGenerating new graph structures is crucial for a wide range o
 f tasks such as drug discovery and protein design.\nParticularly\, diffusi
 on-based models have set benchmarks in the domain of graph generation.\nBu
 ilding upon this success\, there is a growing research interest in explori
 ng how graph diffusion models can be generalized across different contexts
 .\nIn this report\, we first review foundational work in graph diffusion\,
  detailing diffusion models and elaborating on how they are effectively ap
 plied in the graph domain.\nThen\, due to the square space complexity of s
 uch models\, we proceed to introduce a new approach that generalizes curre
 nt models to generate larger graphs and maintains high expressivity throug
 h multiscale generation.\nAfterward\, we explore a study that interprets d
 iffusion models from an optimal transport perspective\, providing an algor
 ithm that extends these models to tackle more complex distribution shift t
 asks. This advancement facilitates a range of potential downstream applica
 tions\nFinally\, we discuss future research directions in graph generation
 \, with a particular emphasis on enhancing the generalization capacity of 
 current models.\n\nBackground papers\n\n	Permutation Invariant Graph Gener
 ation via Score-Based Generative Modeling (Chenhao Niu\, Yang Song\, Jiami
 ng Song\, Shengjia Zhao\, Aditya Grover\, Stefano Ermon)\, https://arxiv.
 org/abs/2003.00638\n	Efficient and Scalable Graph Generation through Itera
 tive Local Expansion (Andreas Bergmeister\, Karolis Martinkus\, Nathanaël
  Perraudin\, Roger Wattenhofer)\, https://arxiv.org/abs/2312.11529\n	Diff
 usion Schrödinger Bridge with Applications to Score-Based Generative Mode
 ling (Valentin De Bortoli\, James Thornton\, Jeremy Heng\, Arnaud Doucet)\
 , https://arxiv.org/abs/2106.01357\n\n\n 
LOCATION:ELE 242 https://plan.epfl.ch/?room==ELE%20242
STATUS:CONFIRMED
END:VEVENT
END:VCALENDAR