Sotirios Sabanis

About

I am Professor of Stochastic Analysis and Algorithms at the University of Edinburgh, and I hold the Personal Chair of Statistics and Stochastic Analysis at the National Technical University of Athens.

I read Mathematics at the Aristotle University of Thessaloniki and was awarded my PhD by the University of Strathclyde, in the Department of Statistics and Modelling Science. At Edinburgh I served as founding Programme Director for a suite of postgraduate programmes in computational mathematical finance, and as Director of Knowledge Exchange for the School of Mathematics until 2025.

I direct the Centre for Investing Innovation, a strategic partnership between the global asset manager abrdn and the University of Edinburgh, based at the Edinburgh Futures Institute. Much of my research puts rigorous stochastic analysis to work on the problems firms face in practice.

I am an affiliated researcher at the Archimedes Unit of the Athena Research Centre, which works on artificial intelligence, data science and algorithms. I was formerly a Turing Fellow at The Alan Turing Institute.

My work appears in leading journals, among them the Journal of Machine Learning Research, Annals of Applied Probability, Bernoulli, SIAM Journal on Mathematics of Data Science, SIAM Journal on Numerical Analysis, Stochastic Processes and their Applications and Stochastics and Partial Differential Equations. It also appears at the leading AI conferences, including the International Conference on Machine Learning (ICML) and the Conference on Empirical Methods in Natural Language Processing (EMNLP).

Research

Numerical analysis, probability, and the algorithms underpinning modern AI

My focus is explicit numerical algorithms for nonlinear random systems of typically high dimension, and their interplay with core data science, machine learning and AI techniques. Four strands run through the work.

• Learning

  Stochastic optimizers and generative models

  Convergence theory for the optimizers used to train neural networks, and Wasserstein-distance guarantees for score-based generative models under semiconvexity and discontinuous gradients.

• Sampling

  Langevin dynamics and MCMC in high dimension

  Non-asymptotic guarantees for Langevin-type samplers when the target is neither smooth nor log-concave, including performance of the unadjusted algorithm under isoperimetric conditions alone.

• Numerics

  Explicit schemes for SDEs with superlinear coefficients

  Standard Euler methods diverge when drift and diffusion coefficients grow faster than linearly. Tamed schemes recover strong convergence while staying explicit — and therefore cheap — with rates that hold under progressively weaker regularity.

• Finance

  Data-driven modelling for investment

  Applying these methods to financial data: tail risk in high-dimensional portfolios, and AI systems that support investment research and decision-making.

Full research fingerprint on the Edinburgh Research Explorer.

Publications

Recent work

Complete list: Google Scholar · ORCID · Research Explorer

Projects & industry

Funded research and knowledge exchange

• Centre for Investing Innovation

  Director · abrdn and the University of Edinburgh · £7.5m over five years, hosted at the Edinburgh Futures Institute

• AI innovation in consumer packaged-goods supply chains

  Principal Investigator · Innovate UK · 2023–2026 · diffusion-based optimisation and generative models for object recognition in retail execution and smart factories

• The abrdn investment co-pilot: statistical intelligence at work

  Co-investigator · abrdn · 2024–2026

• Portfolio dimensionality and data-driven tail risk management

  Principal Investigator, with J. Gondzio · industry funded · 2019–2020

• TRAIN@Ed

  Principal Investigator (MATHS strand) and co-investigator · European Commission · 2019–2023

My research portfolio includes funding from the Alan Turing Institute (including a Fellowship), the Royal Society, EPSRC and Innovate UK.

Teaching & students

Doctoral supervision and programme design

Research group

• Nikolaos Makras PhD student, University of Edinburgh
• Tim Johnston Former PhD student · now postdoctoral researcher at CEREMADE, Université Paris-Dauphine – PSL
• Iosif Lytras Former PhD student · now postdoctoral researcher at the Archimedes Unit, Athena Research Centre
• Ying Zhang Former PhD student · now Assistant Professor at the Hong Kong University of Science and Technology (Guangzhou)
• Stefano Bruno Former postdoctoral researcher · now at UNIST, South Korea, and Visiting Researcher at Edinburgh
• Dong-Young Lim Former postdoctoral researcher (Marie Skłodowska-Curie Fellow) · now Assistant Professor in Industrial Engineering and the Artificial Intelligence Graduate School at UNIST, South Korea

I am currently accepting PhD students. If you are interested in numerical methods for SDEs, sampling algorithms, or the theory behind stochastic optimisation and generative models, please get in touch with a short description of your background and interests.

Programme direction

• MSc Computational Mathematical Financefounding Programme Director, 2015–18
• MSc Financial Modelling and Optimizationfounding Programme Director, 2010–18

Talks

Selected invited talks and events organised

• SDEs/SPDEs: theory, numerics and their interplay with data science

  Organiser · University of Crete, Greece

• Analysis of adaptive stochastic gradient and MCMC algorithms

  Organiser · The Alan Turing Institute, London

• MCMC and diffusion techniques

  Organiser · The Alan Turing Institute, London

• Stochastic Analysis in honour of István Gyöngy’s 65th birthday

  Organiser · University of Edinburgh

• International conferences on Monte Carlo and quasi-Monte Carlo methods

  Invited speaker and session chair · Sydney, Rennes, Montréal, Stanford

Contact

Edinburgh