Title:	Postdoc on Graph Neural Networks
DNr:	Berzelius-2024-301
Project Type:	LiU Berzelius
Principal Investigator:	Roman Bresson <bresson@kth.se>
Affiliation:	Kungliga Tekniska högskolan
Duration:	2024-08-16 – 2025-03-01
Classification:	10201
Keywords:

Abstract

I currently work as a postdoc at KTH, studying graph generating models, in particular through multimodal representation of text and graph data, broadening the bridge between LLMs and GNNs, primarily for biomedical applications. My current work aims at comparing different methods for GNNs aggregation function, in particular regarding hierarchical representation and the use of Kolmogorov-Arnold networks. A next step would be evaluating such models on physics simulation and breast cancer classification.