Seminar Information
Thursdays, 11am AP&M 6402
University of California, San Diego
La Jolla, CA 920930112
Map
Organizer: Todd Kemp
Email: tkemp@math.ucsd.edu
Previous years' webpages:
20172018
20162017
20152016
20142015
20132014
20122013
20112012
20102011
20092010
20082009
20072008
20062007
20052006
20042005
20032004
20022003
20002001
19992000
19981999


Fall 2018 Schedule
Sept 27   Nikhil Srivastava, UC Berkeley
 
Concentration for Sums of Random Matrices with Markov Dependence
There are many wellknown concentration results for sums of independent random matrices, e.g. those of Rudelson, AhlswedeWinter, Tropp, and Oliveira.
We move beyond the independent setting, and prove a Chernofftype bound for sums of matrixvalued random variables sampled via a random walk on an
reversible Markov chain, confirming a conjecture due to Wigderson and Xiao. Our proof is based on a new multimatrix extension of the GoldenThompson
inequality which follows from complex interpolation methods.
(Joint work with A. Garg, Y. Lee, and Z. Song.)



Oct 25   LiCheng Tsai, Columbia University
 
Lowertail large deviations of the KPZ equation
Regarding time as a scaling parameter, we prove the onepoint, lower tail Large Deviation Principle (LDP) of the KPZ equation, with an explicit rate function.
This result confirms existing physics predictions. We utilize a formula from [Borodin Gorin 16] to convert LDP of the KPZ equation to calculating an exponential
moment of the Airy point process, and analyze the latter via stochastic Airy operator and Riccati transform.



Nov 8   Joshua Frisch, Caltech
 
Proximal actions, Strong amenability, and infinite conjugacy class groups.
A topological dynamical system (i.e. a group acting by homeomorphisms on a compact topological space) is said to be proximal
if for any two points $p$ and $q$ we can simultaneously push them together i.e. there is a sequence $g_n$ such that
$\lim g_n(p)=\lim g_n (q)$. In his paper introducing the concept of proximality Glasner noted that whenever $\mathbb{Z}$ acts proximally
that action will have a fixed point. He termed groups with this fixed point property "strongly amenable" and showed that nonamenable
groups are not strongly amenable and virtually nilpotent groups are strongly amenable. In this talk I will discuss recent work precisely
characterizing which (countable) groups are strongly amenable.



Nov 15   Noah Forman, University of Washington
 
The diffusion analogue to a treevalued Markov chain
In '99, David Aldous conjectured that a certain natural "random walk" on the space of binary combinatorial trees
should have a continuum analogue, which would be a diffusion on the GromovHausdorff space of continuum trees.
This talk discusses ongoing work by FPalRizzoloWinkel that has recently verified this conjecture with a
pathwise construction of the diffusion. This construction combines our work on dynamics of certain projections
of the combinatorial treevalued random walk with our previous construction of intervalpartitionvalued diffusions.



"Winter" 2019 Schedule
Jan 10   Tom Kurtz, University of Wisconsin
 
Population models as partial observations of genealogical models
Classical models of biological populations, for example, Markov branching processes, typically model population size and possibly
the distribution of types and/or locations of individuals in the population. The intuition behind these models usually includes
ideas about the relationships among the individuals in the population that cannot be directly recovered from the model.
This loss of information is even greater if one employs large population approximations such as the diffusion approximations popular
in population genetics. "Lookdown" constructions provide representations of population models in terms of countable systems of
particles in which each particle has a "type" which may record both spatial location and genetic type and a "level" which incorporates
the lookdown structure which in turn captures the genealogy of the population. The original population model can then be viewed as the
result of partial observation of the more complex model. We will exploit ideas from filtering of Markov processes to make the idea of
partial observation clear and to justify the lookdown construction.



Jan 17   Yi Sun, Columbia University
 
Gaussian fluctuations for products of random matrices
This talk concerns singular values of $M$fold products of i.i.d. rightunitarily invariant $N\times N$ random matrix ensembles.
As $N\to\infty$, the height function of the Lyapunov exponents converges to a deterministic limit by work of
Voiculescu and NicaSpeicher for $M$ fixed and by work of Newman and IsopiNewman for $M$ tending to infinity with $N$.
In this talk, I will show for a variety of ensembles that fluctuations of these height functions about their mean converge
to explicit Gaussian fields which are logcorrelated for $M$ fixed and have a white noise component for $M$ tending to infinity with $N$.
These ensembles include rectangular Ginibre matrices, truncated Haarrandom unitary matrices, and rightunitarily invariant matrices
with fixed singular values. I will sketch our technique, which derives a central limit theorem for global fluctuations via certain
conditions on the multivariate Bessel generating function, a Laplacetransformlike object associated to the spectral measures
of these matrix products. This is joint work with Vadim Gorin.



Jan 31   Haosui Duanmu, UC Berkeley
 
Nonstandard analysis and its application to Markov processes
Nonstandard analysis, a powerful machinery derived from mathematical logic,
has had many applications in probability theory as well as stochastic processes.
Nonstandard analysis allows construction of a single object  a hyperfinite probability
space  which satisfies all the first order logical properties of a finite probability space,
but which can be simultaneously viewed as a measuretheoretical probability space
via the Loeb construction. As a consequence, the hyperfinite/measure duality has
proven to be particularly useful in porting discrete results into their continuous settings.
In this talk, for every generalstatespace continuoustime Markov process satisfying appropriate
conditions, we construct a hyperfinite Markov process which has all the basic order logical properties
of a finite Markov process to represent it. We show that the mixing time and the hitting time agree with
each other up to some multiplicative constants for discretetime generalstatespace reversible Markov
processes satisfying certain condition. Finally, we show that our result is applicable
to a large class of Gibbs samplers and MetropolisHasting algorithms.



Jan 31   Brian Hall, University of Notre Dame Joint with Functional Analysis
 
Eigenvalues of random matrices in the general linear groups
I will consider random matrices in the general linear group GL(N;C) distributed according to a heat kernel measure.
This may also be described as the distribution of Brownian motion in GL(N;C) starting at the identity. Numerically,
the eigenvalues appear to cluster into a certain domain $\Sigma_t$ as $N$ tends to infinity. A natural candidate for
the limiting eigenvalue distribution is the “Brown measure” of the limiting object, which is Biane’s "free multiplicative Brownian motion."
I will describe recent work with Driver and Kemp in which we compute this Brown measure. The talk will be self contained and will have
lots of pictures.



Feb 28   Hanbaek Lyu, UCLA
 
Stable network observables via dynamic embedding of motifs
We propose a novel framework for constructing and computing various stable network observables. Our approach is based on sampling a random homomorphism
from a small motif of choice into a given network. Integrals of the law of the random homomorphism induces various network observables, which include
wellknown quantities such as homomorphism density and average clustering coefficient. We show that these network observables are stable with respect
to renormalized cut distance between networks. For their efficient computation, we also propose two Markov chain Monte Carlo algorithms and analyze
their convergence and mixing times. We demonstrate how our techniques can be applied to network data analysis, especially for hypothesis testing and
hierarchical clustering, through analyzing both synthetic and real world network data.



March 7 APM 2402   Xin Sun, Columbia University
 
Conformal embedding and percolation on the uniform triangulation
Following Smirnov's proof of Cardy's formula and Schramm's discovery of SLE, a thorough understanding of the scaling limit of
critical percolation on the regular triangular lattice the has been achieved. Smirnorv's proof in fact gives a discrete
approximation of the conformal embedding which we call the Cardy embedding. In this talk I will present a joint project
with Nina Holden where we show that the uniform triangulation under the Cardy embedding converges to the Brownian disk
under the conformal embedding. Moreover, we prove a quenched scaling limit result for critical percolation on uniform triangulations.
Time permitting, I will also explain how this result fits in the the larger picture of random planar maps and Liouville quantum gravity.



Spring 2019 Schedule
Apr 4   Mark Meckes, Case Western University
 
Quenched central limit theorem in a corner growth setting
We consider pointtopoint directed paths in a random environment on the twodimensional integer lattice. For a general independent environment under mild
assumptions we show that the quenched energy of a typical path satisfies a central limit theorem as the mesh of the lattice goes to zero. The proofs rely
on concentration of measure techniques and some combinatorial bounds on families of paths. This is joint work with Christian Gromoll and Leonid Petrov.



Apr 4 1pm   Elizabeth Meckes, Case Western University Department Colloquium
 
Eigenvalues of random matrices: convergence of spectral measures and eigenvalue rigidity
The behavior of the eigenvalues of large random matrices is generally very predictable, on multiple scales. Macroscopically, results like the semicircle
law describe the overall shape of the eigenvalue distributions. Indeed, for many natural ensembles of random matrices, we can describe in great detail
the way the distribution of the eigenvalues converges to some limiting deterministic probability measure. On a microscopic scale, we often see the
phenomenon of eigenvalue rigidity, in which individual eigenvalues concentrate strongly at predicted locations. I will describe some general approaches
to these phenomena, with many examples: Wigner matrices, Wishart matrices, random unitary matrices, truncations of random unitary matrices,
Brownian motion on the unitary group, and others.



Apr 11 4pm   Roman Vershynin, UC Irvine Department Colloquium
 
Mathematics of Deep Learning
Deep learning is a rapidly developing area of machine learning, which uses artificial neural networks to
perform learning tasks. Although mathematical description of neural networks is simple, theoretical explanation
of spectacular performance of deep learning remains elusive. Even the most basic questions about it
remain open. For example, how many different functions can a neural network compute? Jointly with Pierre
Baldi (UCI CS) we discovered a general capacity formula for all fully connected boolean networks. The formula predicts,
counterintuitively, that shallow networks have greater capacity than deep ones. So, mystery remains.



Apr 25   Boguslaw Zegarlinski, Imperial College London
 
Coercive inequalities for Markov generators on nilpotent Lie groups
I will review and present some new results on construction, long time behaviour
and coercive inequalities for Markov semigroups on nilpotent Lie groups.



May 2   Ioana Dumitriu, University of Washington and UCSD


May 23   Mike Cranston, UC Irvine


