Profile Image

Welcome!

I'm Sohir Maskey

PhD Student @ LMU Munich · Mathematical Foundations of Artificial Intelligence.

I am a PhD Student at Ludwig-Maximilians University of Munich, focusing on the theoretical foundations and applications of geometric deep learning. My research encompasses topics such as the generalization abilities and expressivity of graph neural networks, as well as the applications of graph neural ODEs in graph representational learning. My work is supported by the NSF-Simons Research Collaborations, known as MoDL.

Previously, I completed my MS in Mathematics at the Technical University of Berlin, writing my thesis on the transferability of graph neural networks under the guidance of Prof. Dr. Gitta Kutyniok. I also hold a BS in Mathematics from the University of Heidelberg, with a minor in Economics.

Selected Papers

News

Dec 2024

At NeurIPS 2024 in Vancouver, I presented our paper Weisfeiler and Leman Go Loopy, selected as an Oral Presentation.

Nov 2024

I was selected as a top reviewer at NeurIPS 2024. Thank you!

Oct 2024

Together with researchers from TU Wien, the University of Siena, Technion, and RWTH Aachen, we launched the GLOW seminar series— Graph Learning on Wednesdays. This initiative fosters lively discussions on the fundamentals and cutting-edge advances in Graph Machine Learning. Everyone is welcome to join!

Dec 2023

I visited the chair of Prof. Soledad Villar at JHU. I gave a talk about A Fractional Graph Laplacian Approach to Oversmoothing.

Nov 2023

I was selected as a top reviewer at NeurIPS. Thank you for the free registration!

Sep 2023

Our paper on A Fractional Graph Laplacian Approach to Oversmoothing has been accepted as an oral presentation at MLG@ECMLPKDD 2023!

Jul 2022

I participated in a research project on p-Laplacian based positional encodings for GNNs at the London Geometry and Machine Learning Summer School. The paper was accepted to the NeurReps workshop at NeurIPS 2022.

Dec 2021

I participated in the Workshop on Interpretability, safety, and security in AI at Isaac Newton Institute for Mathematical Sciences, University of Cambridge.

Nov 2021

I participated in the Workshop on Deep learning and partial differential equations at Isaac Newton Institute for Mathematical Sciences, University of Cambridge.

Contact

Feel free to reach out via email or connect on social media.