The Harnack inequality plays a crucial role in elliptic and parabolic PDEs. In particular, one can characterize ancient positive solutions to parabolic PDEs by using the Harnack inequality. In this talk, we consider the mean curvature flow, a parabolic PDE of hypersurfaces. To study its stability, it is important to show the uniqueness of ancient flows staying in an one-side of self-similarly shrinking flows. After rescaling the ancient one-sided flow converges to the static self-similar solution, and so it is the graph of an evolving positive function defined over the self-similar solution. Then, the positive function is a solution to a parabolic PDE, and we can show the uniqueness by using the Harnack inequality.

We study logarithmic spiraling solutions to the 2d incompressible Euler equations which solve a nonlinear transport system on the unit circle. We show that this system is locally well-posed for L^p data as well as for atomic measures, that is logarithmic spiral vortex sheets. We prove global well-posedness for almost bounded logarithmic spirals and give a complete characterization of the long time behavior of logarithmic spirals. This is due to the observation that the local circulation of the vorticity around the origin is a strictly monotone quantity of time. We are then able to show a dichotomy in the long time behavior, solutions either blow up (in finite or infinite time) or completely homogenize. In particular, bounded logarithmic spirals converge to constant steady states. For vortex logarithmic spiral sheets the dichotomy is shown to be even more drastic where only finite time blow up or complete homogenization of the fluid can and does occur.

그래프 신경망 (GNN: graph neural network)은 그래프에서 높은 표현 능력과 함께 특징 정보를 추출하는 방법론으로 학계와 산업체에서 최근 폭발적인 관심을 받고 있다. 본 세미나에서는 그래프 신경망의 개요 및 주요 동작 원리를 다룬다. 구체적으로, message passing의 원리를 이해하고 state-of-the-art 알고리즘에서 사용한 다양한 message passing 함수를 소개한다. 그래프 신경망을 활용한 다양한 downstream 응용 문제들이 존재하지만, 본 세미나에서는 근본적인 그래프 마이닝 문제 중 하나인 네트워크 정렬 (network alignment)으로의 적용을 다룬다. 네트워크 정렬 문제를 정의하고, 기존 연구 결과물들을 요약하고 한계점에 대해 설명한다. 이를 바탕으로 발표자 연구실에서 제안한 그래프 신경망을 활용한 새로운 점진적 네트워크 정렬 방법을 소개한다. 마지막으로, 그래프 신경망을 사용해 해결할 수 있는 다양한 실세계 응용 문제를 공유하고 토의한다.

We present new (mostly determinantal) expressions for various eigenvalue statistics in random matrix theory. Whenever the eigenvalue $n$-point correlation function is given in terms of $n \times n$ determinants with some kernel, we propose a new kernel that gives the $n$-point correlation function of the eigenvalues conditioned on the event of some eigenvalues already existing at fixed positions. Using such new kernels we obtain determinantal expressions for the joint densities of the $k$ largest eigenvalues, probability density function of the $k$-th largest eigenvalue, density of the first eigenvalue spacing, and many more. Our formulae is highly amenable to numerical computation through the method proposed by Bornemann (2008).

Since conformal field theory (CFT) was introduced to relate various critical statistical physics models to Virasoro representation theory, it has been applied to string theory, condensed matter physics, vertex operator algebra, and probability theory. In this talk, I will explain how the boundary CFT of central charges c (less than or equal to 1) relates to the boundary CFT of 26 - c in terms of boundary conditions. (On the algebraic side, Feigin and Fuchs described the duality between the categories of Verma modules with central charges c and 26 - c to explain the appearance of the critical dimension 26 of the bosonic string theory.) I will present the connection between the boundary CFT and the theory of Schramm-Loewner evolution in various conformal types.