Yonatan-Sverdlov

About Publications

Yonatan Sverdlov

PhD Student / Researcher

I am a PhD student under Prof. Nadav Dym, focusing on unstructured data, Graph Neural Networks (GNNs), and representation-based learning, with particular interest in over-squashing, over-smoothing, and the limitations of MPNNs and GNNs. My research explores separation quality, continuity, Lipschitzness, and their role in model stability. Additionally, I work on point cloud analysis, studying how geometric and topological structures can be effectively processed using deep learning. My goal is to enhance the expressivity, stability, and generalization of GNNs.

Publications

FSW-GNN: A Bi-Lipschitz WL-Equivalent Graph Neural Network
Yonatan Sverdlov, Yair Davidson, Nadav Dym, Tal Amir
arXiv | Code | BibTeX

On the Expressive Power of Sparse Geometric MPNNs
Yonatan Sverdlov, Nadav Dym
arXiv | Code | BibTeX
International Conference on Learning Representations (ICLR) 2025

Revisiting Multi-Permutation Equivariance through the Lens of Irreducible Representations
Yonatan Sverdlov, Ido Springer, Nadav Dym
arXiv | Code | BibTeX
International Conference on Learning Representations (ICLR) 2025