I am a fourth-year PhD student at EPFL, Switzerland interested in inference problems on graphs and structure learning. In particular, I am excited about learning graphical parameters from dynamics without the need for full recovery of all edges.
I am advised by Professor Patrick Thiran at the Information and Network Dynamics Lab (INDY).
Before coming to Switzerland, I obtained my Bachelor with Honours in Computer Science and Physics from the University of Edinburgh, UK, finishing with a First Class Degree.
When I am not working, I enjoy hiking, rowing (very much a beginner) and making things out of paper (above).
Apart from English, I speak German (natively) and French (intermediate, if you speak slowly ;)
Publications
Reducing Sensor Requirements by Relaxing the Network Metric Dimension
PM, Robin Jaccard, Maximilien Dreveton, Aryan Alavi Razavi Ravari, Patrick Thiran
Presented at SIGMETRICS 2025, Stony Brook, NY, USA.
We show that you can localise a source on a network using far fewer sensors if you allow nodes that are very close to each other to be indistinguishable. By relaxing the classic metric dimension requirement, we trade a bit of precision for a large reduction in sensing cost, especially on trees. We also propose a practical two-stage method that first narrows down a region and then pinpoints the source.
Learn to Vaccinate: Combining Structure Learning and Effective Vaccination for Epidemic and Outbreak Control
Sepehr Elahi, PM, Patrick Thiran
Poster Presented at ICML 2025, Vancouver, Canada.
We propose a novel framework that combines structure learning and vaccination strategies to effectively control epidemics on networks with unknown structures. Our approach leverages limited observations of infection spread to learn the underlying network and identify key nodes for vaccination.