The Transport and Mobility Laboratory, directed by Michel Bierlaire, is active in operations research and optimization, with a particular emphasis on transportation systems. We are particularly interested in introducing mathematical models of behavior in optimization frameworks.
In the Chair of Continuous Optimization (OPTIM), we study the theory and applications of optimization. Our current focus is on geometry and non-convexity. Applications of interest connect with computational sciences, machine learning, statistics and robotics.
At DOLA, our goal is to understand and develop the algorithms used for machine learning problems, and in particular those with many “parameters”: training a neural network with back-propagation, searching saddle points for adversarial training, learning a non-parametric generative model, etc. We are often interested in the “infinite overparameterization limit” which enable the use of tools from mathematical analysis (such as the mean-field theory from mathematics physics) and leads to qualitative insights, quantitative guarantees and practical recommendations.
Discrete optimization problems are ubiquitous. Finding an optimal way to route information through a network or determining an energy-efficient way to display an image are just two examples from everyday life. Our research focuses on the design and analysis of algorithms for discrete optimization problems.
Our focus is on advancing fundamental understanding of multi-agent decision-making in uncertain and dynamic environments. Towards this vision, we develop game theory, distributed control, stochastic and data-driven safe control. Our theoretical work is motivated by applications ranging from transportation and power grid systems to rescue robotics.
I am interested in high-dimensional problems, with motivations coming from machine learning, signal processing or combinatorics, and in particular to phase transitions in optimisation and sampling algorithms.
Daniel Kuhn is the director of the Risk Analytics and Optimization (RAO) lab. Research at RAO addresses the theoretical foundations and applications of optimization under uncertainty with a special focus on and data-driven optimization as well as stochastic, robust, and distributionally robust optimization. Application areas range from statistics and machine learning to business analytics, risk management and engineering.
I am interested in combinatorial optimization on random structures, and continuous optimization in random non-convex landscapes, with motivations coming from the theory of neural networks, inverse problems in signal processing or combinatorics.