学术快报
讲座丨汇总丨报告丨整理
第五十四期
11月10日至11月16日
同济、复旦、上交数院及上海数学与交叉学科研究院学术报告汇总
上海数学与交叉学科研究院学术报告
请至以下公众号查收
01
Sharp boundary regularity properties in hypoelliptic kinetic equations
朱聿哲
(University of Chicago)
2025-11-10 14:00-15:00
会议室(703)
The interplay between transport and degenerate diffusion in kinetic equations gives rise to hypoelliptic smoothing effect. Yet singular boundary sets inherent to kinetic transport flows impede the gain of regularity up to the boundary from reaching a level comparable to that achieved in the interior. We will discuss recent advances on boundary regularity for Fokker-Planck-Landau-type kinetic models, highlighting sharp boundary behaviours of solutions.
上海交通大学数学科学学院
https://math.sjtu.edu.cn/Default/seminarshow/tags/MDAwMDAwMDAwMLJ4epSGyqF2
02
Transfer DAG Learning and Its Application to Biological Regulatory Networks
任明旸 副教授
（上海交通大学）
2025-11-10 16:00-17:30
致远楼101室
Directed acyclic graph (DAG) is pivotal for modeling biological regulatory networks, such as brain functional connectivity networks and gene regulatory networks, yet data from a single study is often limited for accurate DAG reconstruction. Although multi-source datasets provide augmented sample sizes, substantial heterogeneity persists across real-world cohorts. It remains an open question how to pool the heterogeneous data with only local similarity together for better DAG structure reconstruction in the target study. In this work, we introduce a novel set of structural similarity measures for DAGs and then present a transfer DAG learning framework by effectively leveraging information from auxiliary DAGs with node-level local similarities. Our theoretical analysis shows substantial improvement in terms of DAG reconstruction in the target study, which is in sharp contrast to most existing transfer learning methods requiring global similarity. The advantage of the proposed transfer DAG learning is also supported by extensive numerical experiments on both synthetic data and multi-source brain imaging and cancer genomics data.
同济大学数学科学学院
https://math.tongji.edu.cn/info/1385/13133.ht
03
Bakry-Emery Ricci Curvature and Volume Comparison
朱小华 教授
（北京大学）
2025-11-11 9:30-11:20
致远楼108室
We will talk about various volume comparison theory via Bakry-Emery Ricci curvature （BB curvature）. Analogs of Ricci curvature, BB curvature palys a crucial role in many geometric topics， such as conformal geometry, Ricci soliton, Ricci flow, etc. Some application of volume comparison results will be also discussed in the study of compactness of Riemannian metrics.
同济大学数学科学学院
https://math.tongji.edu.cn/info/1385/13138.htm
04
Optimal decay of eigenvector overlap for non-Hermitian random matrices
许媛媛
(中科院应用数学所)
2025-11-12 14:00 — 15:00
光华东主楼1801室
We consider the standard overlap of any bi-orthogonal family of left and right eigenvectors of a large random matrix with centred i.i.d. entries and we prove that it decays as an inverse second power of the distance between the corresponding eigenvalues. This extends similar results for the complex Gaussian ensemble from Bourgade and Dubach, as well as Benaych-Georges and Zeitouni, to any i.i.d. matrix ensemble in both symmetry classes. Based on a joint work with Giorgio Cipolloni and Laszlo Erdos.
复旦大学数学科学学院
https://math.fudan.edu.cn/85/40/c30472a755008/page.htm
05
Escape of mass for higher rank diagonal actions
Wooyeon Kim
(Korea Institute For Advanced Study)
2025-11-12 14:00
Room 102, SCMS
We discuss the escape of mass for higher-rank diagonal actions on the space of lattices and its applications to the inhomogeneous version of the Littlewood conjecture. We first show that the Hausdorff dimension of the set of points that are A-divergent on average in the (d−1)-dimensional closed horosphere in the space of d-dimensional Euclidean lattices, where A is the group of positive diagonal matrices, is at most (d−1)/2. Using these dimension estimates on mass and entropy escape for higher-rank diagonal actions, we then compute the Hausdorff dimension of the exceptional set for the inhomogeneous uniform version of the Littlewood conjecture.
复旦大学数学科学学院
https://math.fudan.edu.cn/85/46/c30472a755014/page.htm
06
Learning Robust Decision Rules for Censored and Confounded Data
崔逸凡
（浙江大学）
2025-11-12 14:00-15:00
会议室(706)
In this talk, we propose two robust criteria for learning optimal treatment rules with censored survival outcomes. The first one aims to identify a treatment rule that maximizes the restricted mean survival time, where the restriction is specified by a given quantile such as the median; the second one focuses on maximizing buffered survival probabilities, with the threshold adaptively adjusted to account for the restricted mean survival time. Moreover, we develop robust treatment rules that enable reliable policy recommendations when unmeasured confounding is present, using the proximal causal inference framework. Simulation studies and real-world applications demonstrate the superior performance of the proposed methods.
上海交通大学数学科学学院
https://math.sjtu.edu.cn/Default/seminarshow/tags/MDAwMDAwMDAwMLJ4epSF2qF2
07
几何不变量在凯勒几何中的应用
朱小华 教授
(北京大学)
2025-11-12 18:25-20:15
同心楼224室
近三, 四十年来，在研究凯勒-爱因斯坦方程，凯勒-里奇流等重要问题上，一些新的几何不变量，像田刚的\alpha-不变量, 丁伟岳-田刚-Futaki不变量，以及后来我们发展的H-不变量等，起着至关重要的作用。这些不变量把凯勒几何，复分析，代数几何等之间联系构建起桥梁。我们的报告内容包括这些不变量的形成历史，所起的作用以及将来的进一步的发展等。
同济大学数学科学学院
https://math.tongji.edu.cn/info/1385/13139.htm
08
A multilayer probit network model for community detection with dependent edges and layers
Dapeng Shi
(Chinese University of Hong Kong)
2025-11-13 14:00-15:30
SCMS Room 102
Community detection in multilayer networks, which aims to identify groups of nodesexhibiting similar connectivity patterns across multiple network layers, has attractedconsiderable attention in recent years. Most existing methods are based on theassumption that edges within the same layer are dependent, or that the layers areeither independent or follow specific dependence structures. In this article, wepropose a novel method for community detection in multilayer networks that accounts for a broad range of inter-layer dependence structures, as well as edge dependence.The proposed method integrates the multilayer stochastic block model for communitydetection with a multivariate probit model to capture dependence structures bothbetween layers and within the same layer. To facilitate the parameter estimation, wedevelop a constrained pairwise likelihood method coupled with an efficientalternating updating algorithm. The asymptotic properties of the proposed method are also established, with a focus on examining the influence of edge dependence, as wellas inter-layer dependence structures and strength on parameter estimation andcommunity detection. The theoretical results are supported by extensive numericalexperiments on both simulated networks and a real-world multilayer trade network.
复旦大学数学科学学院
https://math.fudan.edu.cn/85/43/c30472a755011/page.htm
09
What Makes Looped Transformers Perform Better Than Non-Recursive Ones (Provably)
滕佳烨
(上海财经大学)
2025-11-13 14:00-15:00
光华东主楼1513
While looped transformers (termed as Looped-Attn) often outperform standard transformers (termed as Single-Attn) on complex reasoning tasks, the theoretical basis for this advantage remains underexplored. In this paper, we explain this phenomenon through the lens of loss landscape geometry, inspired by empirical observations of their distinct dynamics at both sample and Hessian levels. To formalize this, we extend the River-Valley landscape model by distinguishing between U-shaped valleys (flat) and V-shaped valleys (steep). Based on empirical observations, we conjecture that the recursive architecture of Looped-Attn induces a landscape-level inductive bias towards River-V-Valley. Theoretical derivations based on this inductive bias guarantee a better loss convergence along the river due to valley hopping, and further encourage learning about complex patterns compared to the River-U-Valley induced by Single-Attn. Building on this insight, we propose SHIFT (Staged HIerarchical Framework for Progressive Training), a staged training framework that accelerates the training process of Looped-Attn while achieving comparable performances.
复旦大学数学科学学院
https://math.fudan.edu.cn/85/dd/c30472a755165/page.htm
10
On Boundary Value Problems for Special Lagrangian Type Equations
邱国寰 副研究员
(中科院数学研究所)
2025-11-13 14:00-15:30
致远楼101室
Special Lagrangian type equations form a class of geometric partial differential equations that characterize special Lagrangian submanifolds—minimal Lagrangian graphs in Euclidean space. These equations arise naturally in calibrated geometry and mirror symmetry. We will discuss Dirichlet, Neumann, and the second boundary value problems for these equations.
同济大学数学科学学院
https://math.tongji.edu.cn/info/1385/13052.htm
11
潮汐力效应、爱因斯坦方程与引力波
谢纳庆 教授
(复旦大学)
2025-11-13 15:30-16:30
致远楼108室
1915年11月25日，爱因斯坦在普鲁士科学院的第四次演讲中提出了著名的“引力场方程”。广义相对论是关于引力的一项革命性理论，其提出彻底改变了我们对世界与宇宙的认识。在这个科普报告中，我们将介绍一种通过潮汐力效应来给出猜测爱因斯坦引力场方程的方法，引力场方程的初值问题以及描述引力波的基本数学框架。我们假设听众具备（多元）微积分与大学物理的基本知识。
同济大学数学科学学院
https://math.tongji.edu.cn/info/1385/13132.htm
12
Spectral theory of free group representations
Jean-Francois Quint
(CNRS & Montpellier University)
2025-11-14 9:30-10:30
Room 102, SCMS
In this talk, I will define a new family of unitary representations of (virtually) free groups. The interest of these representations lies in the fact that for them, certain spectral invariants may be computed explicitely.
复旦大学数学科学学院
https://math.fudan.edu.cn/85/48/c30472a755016/page.htm
13
A conditioned local limit theorem via heat kernel approximation
Ion Grama
(Universite Bretagne-Sud)
2025-11-14 11:00-12:00
Room 102, SCMS
We study a real-valued random walk with centered increments and a finite (2+delta)-moment. Using a previously established approximation of the Gaussian heat kernel for persistence probabilities, we prove a new local limit theorem conditioned by the exit time from the half-line. This result unifies and generalizes existing theorems and provides uniform asymptotics with respect to both the starting and ending points.
复旦大学数学科学学院
https://math.fudan.edu.cn/85/47/c30472a755015/page.htm
14
Almost elementary groupoids and classifiable c*-algebras
Jianchao Wu
(Shanghai Center for Mathematical Sciences)
2025-11-14 16:00-17:00
Room 102, SCMS
This talk is based on collaborations with Xin Ma, where we introduce and study a notion of almost elementary étale groupoids (possibly non-ample), which unifies the notions of almost finiteness and pure infiniteness, and implies strict comparison of the groupoid. Our main motivations lie in the connections to the theory of C*-algebras. Most importantly, we show that an almost elementary minimal étale groupoid gives rise to a tracially Z-stable C*-algebra. On the other hand, we verify that many classifiable C*-algebras (including all of the strongly self-absorbing ones) have almost elementary étale groupoid models. We also make progress toward the conjecture that the Jiang-Su algebra Z does not arise from a transformation groupoid.
复旦大学数学科学学院
https://math.fudan.edu.cn/85/44/c30472a755012/page.htm
15
Deformations, cohomologies and homotopies of Rota-Baxter Lie algebras
生云鹤 教授
(吉林大学)
2025-11-14 14:00-15:00
会议室(706)
We determine the L-infty-algebra that controls deformations of a relative Rota-Baxter Lie algebra and show that it is an extension of the dg Lie algebra controlling deformations of the underlying LieRep pair by the dg Lie algebra controlling deformations of the relative Rota-Baxter operator. Consequently, we define the cohomology of relative Rota-Baxter Lie algebras and relate it to their infinitesimal deformations. The notion of a homotopy relative Rota-Baxter Lie algebra is introduced. We show that a class of homotopy relative Rota-Baxter Lie algebras is intimately related to pre-Lie-infty algebras. This talk is based on joint works with Chengming Bai, Li Guo, Andrey Lazarev and Rong Tang.
上海交通大学数学科学学院
https://math.sjtu.edu.cn/Default/seminarshow/tags/MDAwMDAwMDAwMLJ4et6HtKF2
16
A new class of irreducible sl_3-modules
David Ridout
(University of Melbourne)
2025-11-14 15:00-16:00
会议室(706)
The quest to understand the representation theory of affine vertex algebras has renewed interest in some of the less familiar corners of Lie algebra representation theory. Here, I report on work associated to sl_3 which appears to take us beyond what has previously been explored.
上海交通大学数学科学学院
https://math.sjtu.edu.cn/Default/seminarshow/tags/MDAwMDAwMDAwMLJ4epSGtKF2
17
Convexification Numerical Method for a Coefficient Inverse Problem for the Radiative Transport Equation
杨志鹏 副教授
(兰州大学)
2025-11-15 10:00-10:30
致远楼101室
This talk concerns the coefficient inverse problem for the stationary radiative transport equation. Based on Carleman estimate, the globally convergent so-called convexification numerical method is developed and its convergence analysis is provided. For the numerical implementation, an orthonormal basis is used. Extensive numerical studies in the 2-D case are presented. This talk is based on the joint works with M. V. Klibanov, J. Li, L. H. Nguyen, and V. G. Romanov.
同济大学数学科学学院
https://math.tongji.edu.cn/info/1385/13140.htm
18
耦合动态边界条件的热力学一致PDE的建模与计算
景晓波 副研究员
(东南大学)
2025-11-15 10:30-11:00
致远楼101室
偏微分方程与热力学的发展并非相互独立，而是相互关联、相互促进的。这是因为无论偏微分方程还是热力学都需要考虑内部与外界的物质及能量交换。在偏微分方程领域，往往通过设置合适的边界条件来表述交换关系；而在热力学领域，常常通过选取适当的系综来刻画非平衡系统。近年来，众多学者在尝试将两个领域密切联系起来，并提出热力学一致模型这一概念，从而启发出很多建模、分析及数值算法。在该报告中，我们将以相场为例，介绍动态边界条件的历史、报告最新的研究成果并提出一些亟待解决的问题。在热力学一致的准则下，从可计算建模以及保结构计算两方面重点阐明不同边界条件的选取准则，相关关系以及物理意义。
同济大学数学科学学院
https://math.tongji.edu.cn/info/1385/13141.htm
来源丨同济大学数学科学学院研究生会
收集丨张鑫龙
排版丨张鑫龙
审核丨赵紫烨
