Random Matrices with Structured or Unstructured Correlations

Speaker: 

Todd Kemp

Institution: 

UCSD

Time: 

Tuesday, April 17, 2018 - 11:00pm to 11:50pm

Host: 

Location: 

306 RH

Random matrix theory began with the study, by Wigner in the 1950s, of high-dimensional matrices with i.i.d. entries (up to symmetry).  The empirical law of eigenvalues demonstrates two key phenomena: bulk universality (the limit empirical law of eigenvalues doesn't depend on the laws of the entries) and concentration (the convergence is robust and fast).

 

Several papers over the last decade (initiated by Bryc, Dembo, and Jiang in 2006) have studied certain special random matrix ensembles with structured correlations between some entries. The limit laws are different from the Wigner i.i.d. case, but each of these models still demonstrates bulk universality and concentration.

 

In this lecture, I will talk about very recent results of mine and my students on these general phenomena:

 

Bulk universality holds true whenever there are constant-width independent bands, regardless of the correlations within each band.  (Interestingly, the same is not true for independent rows or columns, where universality fails.)  I will show several examples of such correlated band matrices generalizing earlier known special cases, demonstrating how the empirical law of eigenvalues depends on the structure of the correlations.

 

At the same time, I will show that concentration is a more general phenomenon, depending not on the the structure of the correlations but only on the sizes of correlated partition blocks. Under  some regularity assumptions, we find that Gaussian concentration occurs in NxN ensembles so long as the correlated blocks have size smaller than N^2/log(N).

Distribution of descents in matchings

Speaker: 

Gene Kim

Institution: 

USC

Time: 

Tuesday, February 20, 2018 - 11:00am to 12:00pm

Host: 

Location: 

RH 306

The distribution of descents in certain conjugacy classes of S_n have been previously studied, and it is shown that its moments have interesting properties. This paper provides a bijective proof of the symmetry of the descents and major indices of matchings (also known as fixed point free involutions) and uses a generating function approach to prove an asymptotic normality theorem for the number of descents in matchings.

 

Efficient algorithms for phase retrieval in high dimensions

Speaker: 

Yan Shuo Tan

Institution: 

University of Michigan

Time: 

Thursday, February 8, 2018 - 11:00am to 12:00pm

Host: 

Location: 

RH 306P

Mathematical phase retrieval is the problem of solving systems of rank-1 quadratic equations. Over the last few years, there has been much interest in constructing algorithms with provable guarantees. Both theoretically and empirically, the most successful approaches have involved direct optimization of non-convex loss functions. In the first half of this talk, we will discuss how stochastic gradient descent for one of these loss functions provably results in (rapid) linear convergence with high probability. In the second half of the talk, we will discuss a semidefinite programming algorithm that simultaneously makes use of a sparsity prior on the solution vector, while overcoming possible model misspecification.

Bi-free probability and an approach to conjugate variables

Speaker: 

Ian Charlesworth

Institution: 

UCSD

Time: 

Tuesday, February 27, 2018 - 11:00am

Location: 

RH 306

Free entropy theory is an analogue of information theory in a non-commutative setting, which has had great applications to the examination of structural properties of von Neumann algebras. I will discuss some ongoing joint work with Paul Skoufranis to extend this approach to the setting of bi-free probability which attempts to study simultaneously ``left'' and ``right'' non-commutative variables. I will speak in particular of an approach to a bi-free Fisher information and bi-free conjugate variables -- analogues of Fisher's information measure and the score function of information theory. The focus will be on constructing these tools in the non-commutative setting, and time permitting, I will also mention some results such as bi-free Cramer-Rao and Stam inequalities, and some quirks of the bi-free setting which are not present in the free setting.

Phase transition in the spiked random tensors

Speaker: 

Wei-Kuo Chen

Institution: 

University of Minnesota

Time: 

Thursday, January 4, 2018 - 11:00am to 11:50am

Host: 

Location: 

RH 306

The problem of detecting a deformation in a symmetric Gaussian random tensor is concerned about whether there exists a statistical hypothesis test that can reliably distinguish a low-rank random spike from the noise. Recently Lesieur et al. (2017) proved that there exists a critical threshold so that when the signal-to-noise ratio exceeds this critical value, one can distinguish the spiked and unspiked tensors and weakly recover the spike via the minimal mean-square-error method. In this talk, we will show that in the case of the rank-one spike with Rademacher prior, this critical value strictly separates the distinguishability and indistinguishability of the two tensors under the total variation distance. Our approach is based on a subtle analysis of the high temperature behavior of the pure p-spin model, arising initially from the field of spin glasses. In particular, the signal-to-noise criticality is identified as the critical temperature, distinguishing the high and low temperature behavior, of the pure p-spin model.

Asymptotic analysis of multi-class queues with random order of service .

Speaker: 

Reza Aghajani

Institution: 

UCSD

Time: 

Saturday, December 2, 2017 - 3:20pm to 4:10pm

Location: 

NS2 1201

The random order of service (ROS) is a natural scheduling policy for systems where no ordering of customers can or should be established. Queueing models under ROS have been used to study molecular interactions of intracellular components in biology. However, these models often assume exponential distributions for processing and patience times, which is not realistic especially when operations such as binding, folding, transcription and translation are involved. We study a multi-class queueing model operating under ROS with reneging and generally distributed processing and patience times. We use measure-valued processes to describe the dynamic evolution of the network, and establish a fluid approximation for this representation. Obtaining a fluid limit for this network requires a multi-scale analysis of its fast and slow components, and to establish an averaging principle in the context of measure-valued process. In addition, under slightly more restrictive assumptions on the patience time distribution, we introduce a reduced, function-valued fluid model that is described by a system of non-linear Partial Differential Equations (PDEs). These PDEs, however, are non-standard and the analysis of their existence, uniqueness and stability properties requires new techniques.

Harnack inequality for degenerate balanced random random walks.

Speaker: 

Jean-Dominique Deuschel

Institution: 

Technische Universitat, Berlin

Time: 

Saturday, December 2, 2017 - 2:00pm to 2:50pm

Location: 

NS2 1201

We consider an i.i.d. balanced environment  $\omega(x,e)=\omega(x,-e)$, genuinely d dimensional on the lattice and show that there exist a positive constant $C$ and a random radius $R(\omega)$ with streched exponential tail such that every non negative

$\omega$ harmonic function $u$ on the ball  $B_{2r}$ of radius $2r>R(\omega)$,

we have $\max_{B_r} u <= C \min_{B_r} u$.

Our proof relies on a quantitative quenched invariance principle

for the corresponding random walk in  balanced random environment and

a careful analysis of the directed percolation cluster.

This result extends Martins Barlow's Harnack's inequality for i.i.d.

bond percolation to the directed case.

This is joint work with N.Berger  M. Cohen and X. Guo.

On the Navier-Stokes equation with rough transport noise.

Speaker: 

James-Michael Leahy

Institution: 

USC

Time: 

Saturday, December 2, 2017 - 11:20am to 12:10pm

Location: 

NS2 1201

In this talk, we present some results on the existence of weak-solutions of the Navier-Stokes equation perturbed by transport-type rough path noise with periodic boundary conditions in dimensions two and three. The noise is smooth and divergence free in space, but rough in time. We will also discuss the problem of uniqueness in two dimensions. The proof of these results makes use of the theory of unbounded rough drivers developed by M. Gubinelli et al.

 

As a consequence of our results, we obtain a pathwise interpretation of the stochastic Navier-Stokes equation with Brownian and fractional Brownian transport-type noise. A Wong-Zakai theorem and support theorem follow as an immediate corollary. This is joint work with Martina Hofmanov\'a and Torstein Nilssen.

Deviations of random matrices and applications.

Speaker: 

Roman Vershynin

Institution: 

UCI

Time: 

Saturday, December 2, 2017 - 10:00am to 10:50am

Location: 

NS2 1201

Uniform laws of large numbers provide theoretical foundations for statistical learning theory. This lecture will focus on quantitative uniform laws of large numbers for random matrices. A range of illustrations will be given in high dimensional geometry and data science.

Pages

Subscribe to RSS - Combinatorics and Probability