Seminars and Colloquia by Series

Thursday, November 30, 2017 - 11:05 , Location: Skiles 006 , Zhou Fan , Stanford University , zhoufan@stanford.edu , Organizer: Michael Damron
Random effects models are commonly used to measure genetic variance-covariance matrices of quantitative phenotypic traits. The population eigenvalues of these matrices describe the evolutionary response to selection. However, they may be difficult to estimate from limited samples when the number of traits is large. In this talk, I will present several results describing the eigenvalues of classical MANOVA estimators of these matrices, including dispersion of the bulk eigenvalue distribution, bias and aliasing of large "spike" eigenvalues, and distributional limits of eigenvalues at the spectral edges. I will then discuss a new procedure that uses these results to obtain better estimates of the large population eigenvalues when there are many traits, and a Tracy-Widom test for detecting true principal components in these models. The theoretical results extend proof techniques in random matrix theory and free probability, which I will also briefly describe.This is joint work with Iain Johnstone, Yi Sun, Mark Blows, and Emma Hine.
Monday, February 27, 2017 - 15:05 , Location: Skiles 006 , Sabyasachi Chatterjee , University of Chicago , sabyasachi@uchicago.edu , Organizer: Christian Houdre
We consider the problem of estimating pairwise comparison probabilities in a tournament setting after observing every pair of teams play with each other once. We assume the true pairwise probability matrix satisfies a stochastic transitivity condition which is popular in the Social Sciences.This stochastic transitivity condition generalizes the ubiquitous Bradley- Terry model used in the ranking literature. We propose a computationally efficient estimator for this problem, borrowing ideas from recent work on Shape Constrained Regression. We show that the worst case rate of our estimator matches the best known rate for computationally tractable estimators. Additionally we show our estimator enjoys faster rates of convergence for several sub parameter spaces of interest thereby showing automatic adaptivity. We also study the missing data setting where only a fraction of all possible games are observed at random.
Thursday, February 23, 2017 - 11:00 , Location: Skiles 005 , Rajarshi Mukherjee , Department of Statistics, Stanford University , Organizer:
In this talk, I will discuss some examples of sparse signal detection problems in the context of binary outcomes. These will be motivated by examples from next generation sequencing association studies, understanding heterogeneities in large scale networks, and exploring opinion distributions over networks. Moreover, these examples will serve as templates to explore interesting phase transitions present in such studies. In particular, these phase transitions will be aimed at revealing a difference between studies with possibly dependent binary outcomes and Gaussian outcomes. The theoretical developments will be further complemented with numerical results.
Tuesday, January 24, 2017 - 10:00 , Location: Skiles 005 , Shishi Luo , UC Berkeley , shishi.luo@berkeley.edu , Organizer:
Biology is becoming increasingly quantitative, with large genomic datasets being curated at a rapid rate. Sound mathematical modeling as well as data science approaches are both needed to take advantage of these newly available datasets. I will describe two projects that span these approaches. The first is a Markov chain model of naturalselection acting at two scales, motivated by the virulence-transmission tradeoff from pathogen evolution. This stochastic model, under a natural scaling, converges to a nonlinear deterministic system for which we can analytically derive steady-state behavior. This analysis, along with simulations, leads to general properties of selection at two scales. The second project is a bioinformatics pipeline that identifies gene copy number variants, currently a difficult problem in modern genomics. This quantificationof copy number variation in turn generates new mathematical questionsthat require the type of probabilistic modelling used in the first project.
Thursday, January 19, 2017 - 14:00 , Location: Skiles 005 , Wenjing Liao , Johns Hopkins University , liao@math.jhu.edu , Organizer: Sung Ha Kang
High-dimensional data arise in many fields of contemporary  science and introduce new challenges in statistical learning due to the  well-known curse of dimensionality. Many data sets in image analysis and  signal processing are in a high-dimensional space but exhibit a  low-dimensional structure. We are interested in building efficient  representations of these data for the purpose of compression and inference,  and giving performance guarantees that are only cursed by the intrinsic  dimension of data. Specifically, in the setting where a data set in $R^D$  consists of samples from a probability measure concentrated on or near an  unknown $d$-dimensional manifold with $d$ much smaller than $D$, we  consider two sets of problems: low-dimensional geometric approximation to  the manifold and regression of a function on the manifold. In the first  case we construct multiscale low-dimensional empirical approximations to the manifold and give finite-sample performance guarantees. In the second   case we exploit these empirical geometric approximations of the manifold to   construct multiscale approximations to the function. We prove finite-sample guarantees showing that we attain the same learning rates as if the function was defined on a Euclidean domain of dimension $d$. In both cases  our approximations can adapt to the regularity of the manifold or the  function even when this varies at different scales or locations. All  algorithms have complexity $C n\log (n)$ where $n$ is the number of  samples, and the constant $C$ is linear in $D$ and exponential in $d$. 
Thursday, January 12, 2017 - 11:05 , Location: Skiles 006 , Tengyuan Liang , University of Pennsylvania , tengyuan@wharton.upenn.edu , Organizer: Michael Damron
Network data analysis has wide applications in computational social science, computational biology, online social media, and data visualization. For many of these network inference questions, the brute-force (yet statistically optimal) methods involve combinatorial optimization, which is computationally prohibitive when faced with large scale networks. Therefore, it is important to understand the effect on statistical inference when focusing on computationally tractable methods. In this talk, we will discuss three closely related statistical models for different network inference problems. These models answer inference questions on cliques, communities, and ties, respectively. For each particular model, we will describe the statistical model, propose new computationally efficient algorithms, and study the theoretical properties and numerical performance of the algorithms. Further, we will quantify the computational optimality through describing the intrinsic barrier for certain efficient algorithm classes, and investigate the computational-to-statistical gap theoretically. A key feature shared by our studies is that, as the parameters of the model changes, the problems exhibit different phases of computational difficulty.
Thursday, December 1, 2016 - 15:05 , Location: Skiles 005 , Kolyan Ray , Leiden Univ. , k.m.ray@math.leidenuniv.nl , Organizer: Ionel Popescu
Asymptotic equivalence between two statistical models means that they have the same asymptotic (large sample) properties with respect to all decision problems with bounded loss. In nonparametric (infinite-dimensional) statistical models, asymptotic equivalence has been found to be useful since it can allow one to derive certain results by studying simpler models. One of the key results in this area is Nussbaum’s theorem, which states that nonparametric density estimation is asymptotically equivalent to a Gaussian shift model, provided that the densities are smooth enough and uniformly bounded away from zero.We will review the notion of asymptotic equivalence and existing results, before presenting recent work on the extent to which one can relax the assumption of being bounded away from zero. We further derive the optimal (Le Cam) distance between these models, which quantifies how close they are for finite-samples. As an application, we also consider Poisson intensity estimation with low count data. This is joint work with Johannes Schmidt-Hieber.
Wednesday, January 27, 2016 - 11:00 , Location: Skiles 006 , Jacopo De Simoi , Paris Diderot University , jacopods@math.utoronto.ca , Organizer: Federico Bonetto
Dynamical billiards constitute a very natural class of Hamiltonian systems: in 1927 George Birkhoff conjectured that, among all billiards inside smooth planar convex domains, only billiards in ellipses are integrable. In this talk we will prove a version of this conjecture for convex domains that are sufficiently close to an ellipse of small eccentricity. We will also describe some remarkable relation with inverse spectral theory and spectral rigidity of planar convex domains. Our techniques can in fact be fruitfully adapted to prove spectral rigidity among generic (finitely) smooth axially symmetric domains which are sufficiently close to a circle. This gives a partial answer to a question by P. Sarnak.
Thursday, January 21, 2016 - 11:05 , Location: Skiles 006 , Craig Schroeder , UCLA , Organizer: Haomin Zhou
Hybrid particle/grid numerical methods have been around for a long time, andtheir usage is common in some fields, from plasma physics to artist-directedfluids.  I will explore the use of hybrid methods to simulate many differentcomplex phenomena occurring all around you, from wine to shaving foam and fromsand to the snow in Disney's Frozen.  I will also talk about some of thepractical advantages and disadvantages of hybrid methods and how one of theweaknesses that has long plagued them can now be fixed.
Tuesday, January 19, 2016 - 15:05 , Location: Skiles 005 , Yang Ning , Princeton University , yning@princeton.edu , Organizer: Ionel Popescu
We consider the problem of how to control the measures of false scientific discoveries in high-dimensional models. Towards this goal, we focus on the uncertainty assessment for low dimensional components in high-dimensional models. Specifically, we propose a novel decorrelated likelihood based framework to obtain valid p-values for generic penalized M-estimators. Unlike most existing inferential methods which are tailored for individual models, our method provides a general framework for high-dimensional inference and is applicable to a wide variety of applications, including generalized linear models, graphical models, classifications and survival analysis. The proposed method provides optimal tests and confidence intervals. The extensions to general estimating equations are discussed. Finally, we show that the p-values can be combined to control the false discovery rate in multiple hypothesis testing. 

Pages