Calendar

I will talk about work to uncover connections between invariant theory and maximum likelihood estimation. I will describe how norm minimization over a torus orbit is equivalent to maximum likelihood estimation in log-linear models. We will see the role played by polytopes and discuss connections to scaling algorithms. Based on joint work with Carlos Améndola, Kathlén Kohn, and Philipp Reichenbach.

https://harvard.zoom.us/j/91799784675?pwd=MS9LV25DWk9RcmJoRVM0K3RGWkFRdz09

Password: 1251442

Strominger-Yau-Zaslow conjecture is one of the guiding principles in mirror symmetry, which not only predicts the geometric structures of Calabi-Yau manifolds but also provides a recipe for mirror construction. Besides mirror symmetry, the SYZ conjecture itself is the holy grail in geometrical analysis and closely related to the behavior of the Ricci-flat metrics. In this talk, we will explain how SYZ fibrations on log Calabi-Yau surfaces detect the non-standard semi-flat metric which generalized the semi-flat metrics of Greene-Shapere-Vafa-Yau. Furthermore, we will use the SYZ fibration on log Calabi-Yau surfaces to prove the Torelli theorem of gravitational instantons of type ALH^*. This is based on the joint works with T. Collins and A. Jacob.

https://harvard.zoom.us/j/97335783449?pwd=S3U0eVdyODFEdzNaRXVEUTF3R3NwZz09

Theoretical computer science has given rise to new, exciting questions in algebraic geometry and representation theory. In this talkI will focus on the problem of matrix multiplication. It is generally conjectured by computer scientists that as n grows very large, it becomes almost as easy to multiply nxn matrices as it is to add them! After giving a brief history of the problem I will focus on algebraic geometry and representation theory relevant for the problem and conclude by discussing  recent work with A. Conner, A. Harper, and H. Huang.

There are several properties of closed geodesics which are proven using its Hamiltonian formulation, which has no analogue for minimal surfaces. I will talk about some recent progress in proving some of these properties for minimal surfaces.

Zoom Link: https://cuhk.zoom.us/j/94766047171

(Meeting ID: 947 6604 7171; Passcode: 20220209 )

Over the past several years, attention mechanisms (primarily in the form of the Transformer architecture) have revolutionized deep learning, leading to advances in natural language processing, computer vision, code synthesis, protein structure prediction, and beyond. Attention has a remarkable ability to enable the learning of long-range dependencies in diverse modalities of data. And yet, there is at present limited principled understanding of the reasons for its success. In this talk, I’ll explain how attention mechanisms and Transformers work, and then I’ll share the results of a preliminary investigation into whythey work so well. In particular, I’ll discuss an inductive bias of attention that we call sparse variable creation: bounded-norm Transformer layers are capable of representing sparse Boolean functions, with statistical generalization guarantees akin to sparse regression.

https://harvard.zoom.us/j/99651364593?pwd=Q1R0RTMrZ2NZQjg1U1ZOaUYzSE02QT09

Grothendieck’s Section Conjecture posits that the set of rational points on a smooth projective curve Y of genus at least two should be equal to a certain “section set” defined purely in terms of the etale fundamental group of Y. In this talk, I will preview some upcoming work with Jakob Stix in which we prove a partial finiteness result for this section set, thereby giving an unconditional verification of a prediction of the Section
Conjecture for a general curve Y. We do this by adapting the recent p-adic proof of the Mordell Conjecture due to Brian Lawrence and Akshay Venkatesh.

Superstring theory as we know it started from the discovery by Green and Schwarz in 1984 that the perturbative anomalies of heterotic strings miraculously cancel. But the cancellation of global anomalies of heterotic strings remained an open problem for a long time.

In this talk, I would like to report how this issue was finally resolved last year, by combining two developments outside of string theory. Namely, on one hand, the study of topological phases in condensed matter theory has led to our vastly improved understanding of the general form of global anomalies. On the other hand, the study of topological modular forms in algebraic topology allows us to constrain the data of heterotic worldsheet theories greatly, as far as their contributions to the anomalies are concerned. Putting them together, it is possible to show that global anomalies of heterotic strings are always absent.

The talk is based on https://arxiv.org/abs/2103.12211 and https://arxiv.org/abs/2108.13542 , in collaboration with Mayuko Yamashita.

https://harvard.zoom.us/j/977347126
Password: cmsa

A real algebraic variety is the set of points in real Euclidean space that satisfy a system of polynomial equations.  Metric algebraic geometry is the study of properties of real algebraic varieties that depend on a distance metric. In this talk, we introduce metric algebraic geometry through a discussion of Voronoi cells, bottlenecks, and the reach of an algebraic variety. We also show applications to the computational study of the geometry of data with nonlinear models.

Zoom ID: 950 2372 5230 (Password: cmsa)

It is well-established that the Standard Model (SM) of particle physics is based on su(3)Xsu(2)Xu(1) Lie-algebra. What is less appreciated, however, is that SM accommodates a Z_6 1-form global symmetry.  Gauging this symmetry, or a subgroup of it, changes the global structure of the SM gauge group and amounts to summing over sectors of instantons with fractional topological charges. After a brief review of the concept of higher-form symmetries, I will explain the origin of the Z_6 1-form symmetry and construct the explicit fractional-instanton solutions on compact manifolds. The new instantons mediate baryon-number and lepton-number violating processes, which can win over the weak BPST-instanton processes, provided that SM accommodates extra hyper-charged particles above the TeV scale. I will also comment on the cosmological aspects of the new solutions.

https://harvard.zoom.us/j/977347126
Password: cmsa

My lab is interested in the active and adaptive materials that underlie control of cell shape.  This has centered around understanding force transmission and sensing within the actin cytoskeleton.  I will first review our current understanding of the types of active matter that can be constructed by actin polymers.  I will then turn to our recent experiments to understand how Cell shape changes in epithelial tissue.  I will describe the two sources of active stresses within these tissues, one driven by the cell cycle and controlling cell-cell stresses and the other controlled by cell-matrix signaling controlling motility.  I will then briefly describe how we are using optogenetics to locally control active stresses to reveal adaptive and force-sensitive mechanics of the cytoskeletal machinery. Hopefully, I will convince you that recent experimental and theoretical advances make this a very promising time to study this quite complicated form of active matter!

How many rational preperiodic points can a degree d polynomial in Q[x] have? Conjecturally, there is some uniform bound B_d(Q) on the number of such preperiodic points — but how big is B_d(Q)? Using interpolation we can easily find examples with d + 1 rational preperiodic points, but every point beyond that has to be fought for. In this talk I will share some recent work, joint with John R. Doyle, in which we prove that B_d(Q) is at least d + c*log(d) for some constant c and for all sufficiently large d. As a bonus, I’ll show how our construction also produces examples of pairs of degree d polynomials f(x) and g(x) with more than d^2 common complex preperiodic points.

http://people.math.harvard.edu/~demarco/AlgebraicDynamics/ for more information