Mathematical optimization · Oslo
Thomas M. Surowiec
Chief Research Scientist, Simula Research Laboratory Deputy Centre Director, SURE-AI
About
Optimization in infinite dimensions
I develop theory and algorithms for optimization in function spaces, from risk-averse decision-making under uncertainty to structure-preserving finite element methods, and use them to make engineering and AI systems more reliable.
I am Chief Research Scientist in the Department of Scientific Computing and Numerical Analysis at Simula Research Laboratory, where I have also served as Acting Head of Department, and Deputy Centre Director of SURE-AI, the Norwegian Centre for Sustainable, Risk-Averse and Ethical AI.
My research is in the theory and numerical solution of optimization problems governed by partial differential equations and variational inequalities. Recurring themes are nonsmooth and variational analysis in function spaces, optimal control, stochastic optimization in infinite dimensions, and finite element methods that preserve the structure of the underlying problem, such as Proximal Galerkin and the latent variable proximal point method. Applications range from topology optimization and liquid crystals to semiconductors and microfluidics.
Before joining Simula in 2022, I was Professor (W2) of Mathematical Optimization at Philipps-Universität Marburg (2016–2022) and Assistant Professor (W1) of Nonsmooth Optimization and Variational Analysis at Humboldt-Universität zu Berlin (2014–2016).
Highlights
- 2027Plenary lecture at ENUMATH 2027 (upcoming).
- 2026Paper on exponential map models for neural spatial representations accepted at NeurIPS 2026.
- 2026Research Council of Norway grant PRAGMA (2027–2029) on adaptivity and preconditioning for Proximal Galerkin, funding a two-year stay of L. Zikatanov (Penn State) at Simula.
- 2026LaVa, a FRIPRO grant on the latent variable proximal point method, starts (Research Council of Norway, 2026–2029).
- 2025Langtangen Prize, with B. Keith, for “Proximal Galerkin” (Found. Comput. Math.).
- 2025J. T. Oden Faculty Fellow at the Oden Institute, The University of Texas at Austin; plenary lecture at PGMODAYS 2025, Paris-Saclay.
Research themes
More on my research →Optimization under uncertainty and risk
Risk-averse PDE-constrained optimization, stability and sample-based approximation in infinite dimensions.
Structure-preserving numerical methods
Proximal Galerkin, the latent variable proximal point method and SiMPL for topology optimization.
Variational analysis and optimal control
Optimal control of variational inequalities, MPECs and equilibrium problems in function space.
Optimization for trustworthy AI
Risk measures and robust optimization as tools for reliable learning and inference.
Recent work
All publications →- Exponential map models as an interpretable framework for generating neural spatial representations Advances in Neural Information Processing Systems (NeurIPS 2026), to appear (2026) ConferenceDOI Link is to the bioRxiv preprint
- Inference as risk identification Computational Management Science, Vol. 23, No. 2, Art. 34 (2026) DOIarXiv
- Proximal Galerkin: A structure-preserving finite element method for pointwise bound constraints Foundations of Computational Mathematics, Vol. 26, No. 1, pp. 385–481 (2026) Langtangen Prize 2025DOIarXivCode
- Risk-averse optimal control of random elliptic variational inequalities ESAIM: Control, Optimisation and Calculus of Variations, Vol. 31, Art. 71 (2025) DOIarXiv
- The latent variable proximal point algorithm for variational problems with inequality constraints Computer Methods in Applied Mechanics and Engineering, Vol. 445, Art. 118181 (2025) DOIarXivCode
Optimization in Oslo seminar
I founded and organize the OiO seminar, which from 2022 to 2023 brought 23 international speakers in continuous optimization, optimal control and scientific machine learning to Oslo. The series is currently on hiatus; recordings of most talks remain available.