Research

Statistical physics, network science and machine learning for complex systems.

My research develops principled, physics-inspired models that equip data-driven methods with robustness, interpretability and theoretical foundations, especially when applied to highly interconnected and non-stationary environments. Full publication list on Google Scholar.

Themes

Complex networks modelling

Temporal networks, contagion processes, liquidity cascades, and network evolution in financial markets.

Statistical learning via random matrix theory and disordered systems

Extracting signal from noise in high-dimensional data; spectral features of covariance structures.

Machine learning for dynamical systems

Learning under non-stationarity, physics-informed architectures combined with data-driven components, forecasting and risk assessment.

Background

I obtained my PhD in Theoretical Physics from Sapienza University of Rome, working on the statistical physics of disordered systems. My focus has since moved to the intersection of physics, network theory and machine learning, applied especially to economics and finance, collaborating across disciplines on questions of stability, inference and predictability. I am now Associate Professor in the Department of Computer Science at UCL.

Team

Marcelina Marjankowska · PhD student

Part of the EIGENDATA project, working on eigenvector statistics in random matrix ensembles and their application to physics-informed neural networks: finite-size analytical results, numerical simulations and new spectral methods for inference and optimisation.

Kentaro Hoshisashi · PhD 2025

Introduced Whack-a-mole Online Learning (WamOL), a physics-informed approach to real-time implied volatility surface calibration that enforces no-arbitrage and PDE constraints within deep learning models for option pricing.

Workshops and seminars