Calendar

Posted September 30, 2026
Last modified October 8, 2026

Control and Optimization Seminar Questions or comments?

Franco Blanchini, University of Udine, Italy IEEE Fellow, IFAC Fellow
Control Theory and Mechanisms

Mathematics has recently gained recognition as a language for addressing problems in biology, biochemistry, and medicine. Control theory is a branch of mathematics that draws on tools from several mathematical disciplines. Many techniques originally developed to solve control problems can be adapted to address new and relevant questions, both practical and curiosity-driven, in other fields. This talk focuses on the structural analysis of control systems, with the aim of understanding how mechanisms work, why they behave in a certain way, and to what extent they can perform their intended functions reliably, even in the presence of perturbations and disturbances. The first part introduces motivating examples from disciplines outside control theory, illustrating how a control-theoretic approach can provide powerful insights into their underlying principles. The second part introduces the notions of structural and robust properties, discussing paradigmatic case studies from the literature. Robust stability analysis is approached from an inverse perspective: "We know that this system is stable, but why is it so remarkably stable?" Other fundamental concepts, such as (perfect) adaptation, structural steady-state analysis, graph-based loop analysis, and aggregation, are also discussed. The third part presents applications from biology and biochemistry, illustrating the potential impact that control theory, suitably adapted, can have on these disciplines. It highlights how interdisciplinary research can bring fresh perspectives and new challenges to control theorists.

Posted September 21, 2026

Control and Optimization Seminar Questions or comments?

Tim Hoheisel, McGill University, Canada
A Computational Framework for Linear Inverse Problems via the Maximum Entropy on the Mean Method

We present a framework for solving linear inverse problems that is computationally tractable and has mathematical certificates. To this end, we interpret the ground truth of a linear inverse problem as a random vector with unknown distribution. We solve for a distribution which is close to a prior P (guessed or data-driven) measured in the KL-divergence while also having an expectation that yields high fidelity with the given data that defines the problem. After reformulation this yields a strictly convex, finite dimensional optimization problem whose regularizer, the MEM functional, is paired in duality with the log-moment generating function of the prior P. We exploit this computationally via Fenchel-Rockafellar duality. When no obvious guess for P is available, we use data to generate an empirical prior. Using techniques from variational analysis and stochastic optimization, we show that, and at what rate, the solution of the empirical problems converges (as the sample size grows) to the solution of the problem with known prior. Moreover, we show how stochastic subgradient methods can be brought to bear on the MEM dual. This is joint work with Rustum Choksi (McGill), Matthew King-Roskamp (McGill), and Gabriel Rioux (Imperial College).

Posted September 30, 2026

Control and Optimization Seminar Questions or comments?

Daniel Wachsmuth, University of Wuerzburg, Germany
Solution Algorithms for Nonsmooth Optimal Control Problems Based on the Pontryagin Maximum Principle

The talk is motivated by optimal control problems for partial differential equations. The main interest is in problems that involve integral functions of the type \[\int_\Omega g(u(x)) \ dx,\] where $g : \mathbb R \to \mathbb R$ is nonsmooth and nonconvex, and $u$ is the control variable. Examples involve $L^{\scriptscriptstyle r}$-pseudonorms with $r\in [0,1)$ or problems with the constraint $u(x) \in \mathbb Z$ for $x\in \Omega$. Due to the nonconvexity of $g$, the existence of solutions in $L^{\scriptscriptstyle p}$-spaces cannot be proven. Local solutions (if they exist) can be proven to satisfy the Pontryagin maximum principle. A distinctive feature of this result is that no differentiability with respect to the control variable $u$ is required. The maximum principle serves as a starting point to develop an implementable solution algorithm. We prove that the residual in the maximum principle vanishes along the iterates produced by the algorithm. Under additional assumptions, the iterates are a minimizing sequence.

Posted September 21, 2026

Control and Optimization Seminar Questions or comments?

Radu Bot, University of Vienna
TBA

Posted September 22, 2026

Control and Optimization Seminar Questions or comments?

Rene Henrion, Weierstrass Institute for Applied Analysis and Stochastics, Germany
TBA

Posted September 1, 2026

Control and Optimization Seminar Questions or comments?

Jun Liu, University of Waterloo, Canada
TBA