COMPiLE Lab · University of Edinburgh
Mathematical and computational aspects of machine learning and network science
My research lies at the intersection between machine learning and scientific computing. Recent work includes physics-inspired deep learning models for scientific simulations and climate forecasting, machine learning approaches to OOD, deep learning model compression and speed-up, and the theoretical analysis of neural networks in the infinite width and infinite depth limits.
Training and compressing neural networks directly in low-rank form: dynamical low-rank training, geometry-aware fine-tuning and robust, rank-adaptive models.
Related papers →
Nonlinear Perron–Frobenius theory, nonlinear eigenvector problems and the graph p-Laplacian, with algorithms that come with convergence guarantees.
Related papers →
Centrality, core–periphery and community structure in multilayer networks and hypergraphs, and learning on graphs with higher-order interactions.
Related papers →
Neural PDE solvers, physics-inspired architectures and foundation models for scientific simulation and forecasting.
Related papers →
International Conference on Learning Representations (ICLR), 2025
GeoLoRA: Geometric integration for parameter efficient fine-tuning
Advances in Neural Information Processing Systems (NeurIPS), 2022
Low-rank lottery tickets: finding efficient low-rank neural networks via matrix differential equations
SIAM Journal on Mathematics of Data Science, vol. 5, 2023
Core-periphery detection in hypergraphs
Communications Physics, vol. 4:201, 2021
Node and Edge Eigenvector Centrality for Hypergraphs
Interested in joining? See open opportunities →
