Upcoming MAMS Colloquium Series
Fall 2026
10/13/2026, Friday 3:15-4:15 pm in A.W. Smith 349
Speaker: Professor Dmitriy Bilyk – University of Minnesota – Twin Cities
Title: Energy minimization on the sphere and other metric spaces
Abstract: We shall give a brief survey of problems related to the minimization of discrete interaction energies of the form
$$ \sum_{i\neq j} K (x_i, x_j) $$
and energy integrals
$$ \int_\Omega \int_\Omega K(x,y) d\mu (x) d\mu (y).$$
In the first case, the optimization is over all $N$-point configurations $x_1, \dots, x_N$ in a given space $\Omega$, while in the second case, one optimizes over all Borel probability measures $\mu$ on $\Omega$. The latter case can be viewed, both heuristically and rigorously (as well as numerically), as the limit of the discrete energy optimization as the number of points $N$ grows. Intuitively, optimizers can be thought of as equilibrium distributions of $N$ particles or of a continuous unit charge under the interaction given by the kernel $K$, and such problems appear naturally in mathematical physics, discrete and metric geometry, approximation theory, signal processing etc. A particularly interesting case arises when the interaction $K$ is a power of the distance function (Riesz energies/distance integrals/sums of distances). Optimal distributions in this case are intimately connected to the structure and geometry of the underlying metric space. While in many cases one expects uniform equilibrium measures, sometimes they can be lower-dimensional, supported on small sets, or even discrete (clustering). We shall discuss such problems in various geometric contexts (sphere, torus etc), methods employed in these problems stemming from various branches of analysis (potential theory, harmonic and functional analysis, orthogonal polynomials), and connections to problems of discrete geometry, signal processing etc.
10/9/2026, Friday 3:15-4:15 pm in A.W. Smith 349
Speaker: Professor Hwanwoo Kim, UT Dallas
Title: Efficient Gaussian Process Surrogates for Blackbox Optimization and Posterior Approximation
Abstract: In this talk, we explore efficient Gaussian process surrogate modeling in two distinct contexts: bandit optimization and blackbox posterior approximation. For optimization, we propose novel noise-free Bayesian optimization strategies that incorporate a random exploration step to enhance the accuracy of Gaussian process surrogate models. The new algorithms retain the ease of implementation of the classical GP-UCB algorithm, but the additional random exploration step accelerates their convergence, nearly achieving the optimal convergence rate. For blackbox posterior approximation, we propose utilizing optimization iterates as design points to build a Gaussian process surrogate model for the unnormalized log-posterior density when the likelihood is intractable. We show that the Hellinger distance between the true and approximate posterior distributions decays at a near-optimal rate. We demonstrate the effectiveness of our optimization algorithms on benchmark non-convex test functions and in a black-box engineering design problem, and showcase our posterior approximation approach in Bayesian inference for parameters of dynamical systems.
