Youssef Wally
Job description
I focus on advance representation learning techniques for graphs, with a particular emphasis on developing novel similarity measures and clustering methods. My research explores how relationship graphs—extracted from complex datasets such as spatial omics—can capture intricate dependencies beyond pairwise interactions, offering a richer understanding of medical and biological data. Given the challenges posed by varying graph structures and labelled nodes, my work aims to incorporate underlying semantic relationships through knowledge graph embeddings and non-Euclidean geometries. By leveraging these advanced techniques, seeking to develop more meaningful similarity measures that enhance the analysis of medical and healthcare data, ultimately contributing to improved predictive modelling and decision-making in clinical and biomedical applications.
Publications outside Cristin
- DiaMond: Dementia Diagnosis with Multi-Modal Vision Transformers Using MRI and PET
- Personalized k-fold Cross-Validation Analysis with Transfer from Phasic to Tonic Pain Recognition on X-ITE Pain Database
Research interests
Youssef’s interests lie at the intersection of medical and healthcare data, with a strong focus on developing advanced similarity measures for graph-structured data. Specifically, in leveraging representation learning techniques and non-Euclidean embeddings to improve the analysis of patient records, biological networks, and clinical decision-making systems. By incorporating domain knowledge—such as the hierarchical relationships between medical diagnoses, the functional connections in molecular interactions, or the contextual significance of clinical pathways—aiming to create more meaningful similarity measures that enhance predictive modeling and knowledge discovery in healthcare. Additionally, exploring how relational taxonomies of medical concepts and patient trajectories can be effectively embedded to refine data augmentation strategies, ultimately improving the robustness and interpretability of machine learning models in medical applications.