Random matrix perturbations

Speaker: 

Sean O'Rourke

Institution: 

University of Colorado, Boulder

Time: 

Tuesday, May 14, 2019 - 11:00am to 11:50am

Host: 

Location: 

RH 510M

Computing the eigenvalues and eigenvectors of a large matrix is a basic task in high dimensional data analysis with many applications in computer science and statistics. In practice, however, data is often perturbed by noise. A natural question is the following: How much does a small perturbation to the matrix change the eigenvalues and eigenvectors? In this talk, I will consider the case where the perturbation is random. I will discuss perturbation results for the eigenvalues and eigenvectors as well as for the singular values and singular vectors.  This talk is based on joint work with Van Vu, Ke Wang, and Philip Matchett Wood.

Several open problems on the Hamming cube II.

Speaker: 

Paata Ivanisvili

Institution: 

UCI

Time: 

Tuesday, February 5, 2019 - 11:00am to 12:00pm

Host: 

Location: 

306 RH

The Hamming cube of dimension n  consists of vectors of length n with coordinates +1 or -1.  Real-valued functions on the Hamming cube equipped with uniform counting measure can be expressed as Fourier--Walsh series, i.e., multivariate polynomials of n variables +1 or -1. The degree of the function is called the corresponding degree of its multivariate polynomial representation.  Pick a function whose Lp norm is 1. How large the Lp norm of the discrete (graph) Laplacian of the function can be if its degree is at most d, i.e., it lives on ``low frequencies''? Or how small it can be if the function lives on high frequencies, i.e., say all low degree terms (lower than d) are zero? I will sketch some proofs based on joint works (some in progress) with Alexandros Eskenazis.

Longest increasing and decreasing subsequences

Speaker: 

Richard Stanley

Institution: 

University of Miami

Time: 

Tuesday, January 22, 2019 - 11:00am to 12:00pm

Host: 

Location: 

RH 306

An increasing subsequence of a permutation $a_1, a_2,\dots, a_n$ of 
$1,2,\dots, n$ is a subsequence $b_1,b_2,\dots,b_k$ satisfying 
$b_1<b_2<\cdots<b_k$, and similarly for decreasing subsequence. The 
earliest result in this area is due to Erd\H{o}s and Szekeres in 1935: any 
permuation of $1,2,\dots,pq+1$ has an increasing subsequnce of length 
$p+1$ or a decreasing subsequence of length $q+1$. This result turns out 
to be closely connected to the RSK algorithm from the representation 
theory of the symmetric group. A lot of work has been devoted to the 
length $k$ of the longest increasing subsequence of a permutation 
$1,2,\dots,n$, beginning with Ulam's question of determining the average 
value of this number over all such permutations, and culminating with the 
result of Baik-Deift-Johansson on the limiting distribution of this 
length. There are many interesting analogues of longest increasing 
subsequences, such as longest alternating subsequences, i.e., 
subsequences $b_1,b_2,\dots, b_k$ of a permutation $a_1, a_2,\dots, a_n$ 
satisfying $b_1>b_2<b_3>b_4<\cdots$. The limiting distribution of the 
length of the longest alternating subsequence of a random permutation 
behaves very differently from the length of the longest increasing 
subsequence.  We will survey these highlights from the theory of 
increasing and decreasing subsequences.

The Sperner property

Speaker: 

Richard Stanley

Institution: 

MIT

Time: 

Thursday, January 24, 2019 - 4:00pm to 5:00pm

Host: 

Location: 

RH 306

In 1927 Emanuel Sperner proved that if $S_1,\dots,S_m$ are distinct 
subsets of an $n$-element set such that we never have $S_i\subset S_j$, 
then $m\leq \binom{n}{\lfloor n/2\rfloor}$. Moreover, equality is achieved 
by taking all subsets of $S$ with $\lfloor n/2\rfloor$ elements. This 
result spawned a host of generalizations, most conveniently stated in the 
language of partially ordered sets. We will survey some of the highlights 
of this subject, including the use of linear algebra and the cohomology of 
certain complex projective varieties. An application is a proof of a 
conjecture of Erd\H{o}s and Moser, namely, for all integers $n\geq 1$ and 
real numbers $\alpha\geq 0$, the number of subsets with element sum 
$\alpha$ of an $n$-element set of positive real numbers cannot exceed the 
number of subsets of $\{1,2,\dots,n\}$ whose elements sum to $\lfloor 
\frac 12\binom n2\rfloor$. We will conclude by discussing two recent 
proofs of a 1984 conjecture of Anders Bj\"orner on the weak Bruhat order 
of the symmetric group $S_n$.

Random matrix point processes via stochastic processes

Speaker: 

Elliot Paquette

Institution: 

The Ohio State University

Time: 

Thursday, January 10, 2019 - 12:00pm to 1:00pm

Location: 

RH 340P

In 2007, Virág and Válko introduced the Brownian carousel, a dynamical system that describes the eigenvalues of a canonical class of random matrices. This dynamical system can be reduced to a diffusion, the stochastic sine equation, a beautiful probabilistic object requiring no random matrix theory to understand. Many features of the limiting eigenvalue point process, the Sine--beta process, can then be studied via this stochastic process. We will sketch how this stochastic process is connected to eigenvalues of a random matrix and sketch an approach to two questions about the stochastic sine equation: deviations for the counting Sine--beta counting function and a functional central limit theorem.

Based on joint works with Diane Holcomb, Gaultier Lambert, Bálint Vet\H{o}, and Bálint Virág.

Erratic behavior for one-dimensional random walks in a generic quasi-periodic environment

Speaker: 

Maria Saprykina

Institution: 

KTH, Sweden

Time: 

Monday, December 3, 2018 - 11:00am to 12:00pm

Host: 

Location: 

RH 420

Consider a Markov chain on a one-dimensional torus $\mathbb T$, where a moving point jumps from $x$ to $x\pm \alpha$ with probabilities $p(x)$ and $1-p(x)$, respectively, for some fixed function $p\in C^{\infty}(\mathbb T, (0,1))$ and $\alpha\in\mathbb R\setminus \mathbb Q$. Such Markov chains are called random walks in a quasi-periodic environment. It was shown by Ya. Sinai that for Diophantine $\alpha$ the corresponding random walk has an absolutely continuous invariant measure, and the distribution of any point after $n$ steps converges to this measure. Moreover, the Central Limit Theorem with linear drift and variance holds.

In contrast to these results, we show that random walks with a Liouvillian frequency $\alpha$ generically exhibit an erratic statistical behavior. In particular, for a generic $p$, the corresponding random walk does not have an absolutely continuous invariant measure, both drift and variance exhibit wild oscillations (being logarithmic at some times and almost linear at other times), Central Limit Theorem does not hold.

These results are obtained in a joint work with Dmitry Dolgopyat and Bassam Fayad.

Analytic and Numerical Solutions of Traveling Waves of Some Burgers-Type Equations

Speaker: 

Dongming Wei

Institution: 

Nazarbayev University

Time: 

Monday, December 3, 2018 - 4:00pm to 5:00pm

Host: 

Location: 

RH306

In the talk, several Burgers-Type Equations are presented with derivations from Navier- Stokes equations for non-Newtonian fluids. Some of these equations are derived only recently. Analytic solutions are derived for the equations in some cases which generalizes the classical results for Newtonian fluids and numerical solutions are also presented for some more difficult cases. Special functions such as the Gauss hypergeometric function are used in representation s of analytic solutions, nonlinear implicit ODEs are solved numerically to demonstrate the travelling waves. Finally, possible applications in plastic sheet formation are discussed.

An overview: Different data types and education

Speaker: 

Anna Bargagliotti

Institution: 

Loyola Marymount University

Time: 

Thursday, November 29, 2018 - 5:00pm

Location: 

NSII 1201

Data are prevalent in all aspects of society today. Data-driven decision making is important in education, business, politics, psychology, and many more disciplines. In this talk, I will discuss three different data types and research that surrounds them. How can data be visualized affectively? What are statistical techniques for analyzing different data types? How do the analysis techniques vary depending on the research questions one is interested in answering? Through this, I will introduce my different areas of research and also discuss my pathway as an academic.
 

Regularization of probability measures through free addition and convolution semigroups II

Speaker: 

Gregory Zitelli

Institution: 

UCI

Time: 

Thursday, December 6, 2018 - 12:00pm to 1:00pm

Host: 

Location: 

340P

This is a continuation of the talk from last week about the regularizing properties of free additive convolution. We introduce and discuss a variety of tools from noncommutative probability, notably the operations of free, boolean, and monotone additive and multiplicative convolutions, the Belinschi–Nica semigroup, the free divisibility indicator, and complex subordination. The goal of the talk is to provide a broad overview of free probability, including a variety of curious equations and identities, from a complex analytic viewpoint. The talk is based on the works of Belinschi, Bercovici, Nica, Arizmendi, and Hasebe.

Pages

Subscribe to UCI Mathematics RSS