BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//UNC Statistics & Operational Research - ECPv5.6.0//NONSGML v1.0//EN
CALSCALE:GREGORIAN
METHOD:PUBLISH
X-WR-CALNAME:UNC Statistics & Operational Research
X-ORIGINAL-URL:https://stor.unc.edu
X-WR-CALDESC:Events for UNC Statistics & Operational Research
BEGIN:VTIMEZONE
TZID:America/New_York
BEGIN:DAYLIGHT
TZOFFSETFROM:-0500
TZOFFSETTO:-0400
TZNAME:EDT
DTSTART:20200308T070000
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:-0400
TZOFFSETTO:-0500
TZNAME:EST
DTSTART:20201101T060000
END:STANDARD
BEGIN:DAYLIGHT
TZOFFSETFROM:-0500
TZOFFSETTO:-0400
TZNAME:EDT
DTSTART:20210314T070000
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:-0400
TZOFFSETTO:-0500
TZNAME:EST
DTSTART:20211107T060000
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTART;VALUE=DATE:20200805
DTEND;VALUE=DATE:20200806
DTSTAMP:20220128T122708
CREATED:20200610T150056Z
LAST-MODIFIED:20200610T150056Z
UID:9933-1596585600-1596671999@stor.unc.edu
SUMMARY:STOR Boot Camp
DESCRIPTION:Boot camp for
URL:https://stor.unc.edu/event/stor-boot-camp/2020-08-05/
LOCATION:zoom
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=America/New_York:20200805T083000
DTEND;TZID=America/New_York:20200805T133000
DTSTAMP:20220128T122708
CREATED:20200610T150314Z
LAST-MODIFIED:20200610T150314Z
UID:9937-1596616200-1596634200@stor.unc.edu
SUMMARY:CWE Exams
DESCRIPTION:
URL:https://stor.unc.edu/event/cwe-exams-11/2020-08-05/
LOCATION:Hanes Hall\, Hanes Hall\, Chapel Hill\, NC\, 27599\, United States
END:VEVENT
BEGIN:VEVENT
DTSTART;VALUE=DATE:20200806
DTEND;VALUE=DATE:20200807
DTSTAMP:20220128T122708
CREATED:20200610T150056Z
LAST-MODIFIED:20200610T150056Z
UID:9934-1596672000-1596758399@stor.unc.edu
SUMMARY:STOR Boot Camp
DESCRIPTION:Boot camp for
URL:https://stor.unc.edu/event/stor-boot-camp/2020-08-06/
LOCATION:zoom
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=America/New_York:20200806T083000
DTEND;TZID=America/New_York:20200806T133000
DTSTAMP:20220128T122708
CREATED:20200610T150314Z
LAST-MODIFIED:20200610T150314Z
UID:9938-1596702600-1596720600@stor.unc.edu
SUMMARY:CWE Exams
DESCRIPTION:
URL:https://stor.unc.edu/event/cwe-exams-11/2020-08-06/
LOCATION:Hanes Hall\, Hanes Hall\, Chapel Hill\, NC\, 27599\, United States
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=America/New_York:20200806T143000
DTEND;TZID=America/New_York:20200806T160000
DTSTAMP:20220128T122708
CREATED:20200803T150042Z
LAST-MODIFIED:20200803T150042Z
UID:9941-1596724200-1596729600@stor.unc.edu
SUMMARY:Ph.D. Defense: Yuzixuan Zhu
DESCRIPTION:Yuzixuan Zhu \nPreprocessing and First-Order Primal-Dual Algorithms for \nConvex Optimization \nAdvisors: Dr. Gabor Pataki and Dr. Quoc Tran-Dinh \n \nThis talk focuses on two topics in the field of convex optimization: preprocessing algorithms for semidefinite programs (SDPs)\, and first-order primal-dual algorithms for convex-concave saddle-point problems. \n \nIn the first part of this talk\, we introduce Sieve-SDP\, a simple facial reduction algorithm to preprocess SDPs. Sieve-SDP inspects the constraints of the problem to detect lack of strict feasibility\, deletes redundant rows and columns\, and reduces the size of the variable matrix. It often detects infeasibility. It does not rely on any optimization solver: the only subroutine it needs is Cholesky factorization\, hence it can be implemented in a few lines of code in machine precision. We present extensive computational results on several problem collections from the literature\, with many SDPs coming from polynomial optimization. \n \nIn the second part\, we develop two first-order primal-dual algorithms to solve a class of convex-concave saddle-point problems involving non-bilinear coupling function\, which includes SDP as one of the many special cases. Both algorithms are single-loop and have low per-iteration complexity. Our first algorithm can achieve O(1/k) convergence rates on the duality gap in both ergodic (averaging) sense and semi-ergodic sense\, i.e.\, non-ergodic (last-iterate) on the primal\, and ergodic on the dual. This rate can be further improved on non-ergodic primal objective residual using a new parameter update rule. Under strong convexity assumption\, our second algorithm can boost these convergence rates to no slower than O(1/k^2). Our results can be specified to handle general convex cone-constrained problems. We test our algorithms on applications such as two-player games and image processing to compare our algorithms with existing methods. \n
URL:https://stor.unc.edu/event/ph-d-defense-yuzixuan-zhu/
LOCATION:zoom
CATEGORIES:PhD Defense
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=America/New_York:20200807T083000
DTEND;TZID=America/New_York:20200807T133000
DTSTAMP:20220128T122708
CREATED:20200610T150314Z
LAST-MODIFIED:20200610T150314Z
UID:9939-1596789000-1596807000@stor.unc.edu
SUMMARY:CWE Exams
DESCRIPTION:
URL:https://stor.unc.edu/event/cwe-exams-11/2020-08-07/
LOCATION:Hanes Hall\, Hanes Hall\, Chapel Hill\, NC\, 27599\, United States
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=America/New_York:20200911T153000
DTEND;TZID=America/New_York:20200911T163000
DTSTAMP:20220128T122708
CREATED:20200904T150104Z
LAST-MODIFIED:20200904T150104Z
UID:9948-1599838200-1599841800@stor.unc.edu
SUMMARY:Graduate Seminar: Kevin O'Connor
DESCRIPTION:Optimal Transport for Stationary Markov Chains via Policy Iteration \nIn this talk\, we discuss an extension of optimal transport techniques to stationary Markov chains from a computational perspective. In this context\, we show that the standard optimal transport problem does not capture differences in the dynamics of the two chains. Instead\, we study a new problem\, called the optimal transition coupling problem\, in which the optimal transport problem is constrained to the set of stationary Markovian couplings satisfying a certain transition matrix condition. After drawing a connection between this problem and Markov decision processes\, we prove that solutions can be obtained via the policy iteration algorithm. For settings with large state spaces\, we also define a regularized problem\, propose a faster\, approximate algorithm\, and provide bounds on the computational complexity of each iteration. Finally\, we validate our theoretical results empirically\, demonstrating that the approximate algorithm exhibits faster overall runtime with low error in a simulation study. \n \nThis is joint work with Andrew Nobel and Kevin McGoff. \n \nLink to the paper: https://arxiv.org/pdf/2006.07998.pdf \n
URL:https://stor.unc.edu/event/graduate-seminar-kevin-oconnor/
LOCATION:zoom
CATEGORIES:Graduate Seminar
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=America/New_York:20200918T153000
DTEND;TZID=America/New_York:20200918T163000
DTSTAMP:20220128T122708
CREATED:20200916T113016Z
LAST-MODIFIED:20200916T113016Z
UID:9949-1600443000-1600446600@stor.unc.edu
SUMMARY:Graduate Student Seminar
DESCRIPTION:Dr. Shankar Bhamidi\, Director of Graduate Studies\, will host a zoom chat with grad students. Come with questions\, or just check in and enjoy social time.
URL:https://stor.unc.edu/event/graduate-student-seminar/
LOCATION:zoom
CATEGORIES:Graduate Seminar
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=America/New_York:20200921T160000
DTEND;TZID=America/New_York:20200921T171500
DTSTAMP:20220128T122708
CREATED:20200904T090214Z
LAST-MODIFIED:20200904T090214Z
UID:9947-1600704000-1600708500@stor.unc.edu
SUMMARY:STOR Colloquium: Themis Sapsis\, MIT
DESCRIPTION:Output-Weighted Active Sampling for Bayesian Uncertainty Quantification and Prediction of Rare Events \nThemis Sapsis \nWe introduce a class of acquisition functions for sample selection that leads to faster convergence in applications related to Bayesian uncertainty quantification of rare events. The approach follows the paradigm of active learning\, whereby existing samples of a black-box function are utilized to optimize the next most informative sample. The proposed method aims to take advantage of the fact that some input directions of the black-box function have a larger impact on the output than others\, which is important especially for systems exhibiting rare and extreme events. The acquisition functions introduced in this work leverage the properties of the likelihood ratio\, a quantity that acts as a probabilistic sampling weight and guides the active-learning algorithm towards regions of the input space that are deemed most relevant. We demonstrate superiority of the proposed approach in the uncertainty quantification of a hydrological system as well as the probabilistic quantification of rare events in dynamical systems and the identification of their precursors. We also discuss connections and implications for Bayesian optimization and present applications related to path planning for anomaly (rare event) detection in environment exploration. \n \nJoint work with Dr Antoine Blanchard
URL:https://stor.unc.edu/event/stor-colloquium-themis-sapsis-mit/
LOCATION:Hanes Hall\, Hanes Hall\, Chapel Hill\, NC\, 27599\, United States
CATEGORIES:STOR Colloquium
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=America/New_York:20201002T153000
DTEND;TZID=America/New_York:20201002T163000
DTSTAMP:20220128T122708
CREATED:20200925T153200Z
LAST-MODIFIED:20200925T153200Z
UID:9950-1601652600-1601656200@stor.unc.edu
SUMMARY:Graduate Seminar: Quoc Tran-Dinh
DESCRIPTION:Quoc Tran-Dinh\nStatistics & Operations Research Dept.\, UNC-CH \nEfficient Stochastic Gradient-Based Algorithms with Biased Variance-Reduced Estimators \nIn this talk\, we discuss some recent progress in stochastic gradient-based methods using biased variance-reduced estimators to approximate a stationary point or a KKT point of non-convex problems such as stochastic non-convex optimization\, stochastic compositional optimization\, and stochastic minimax problems. More specifically\, we introduce a new class of hybrid biased variance-reduced estimators that combines the well-known SARAH (Nguyen et al (2017)) and the classical SGD estimators to form a new one. We investigate the properties of this estimator class and draw some connections with existing ones. Next\, we develop a new stochastic gradient-based algorithm to solve a class of composite nonconvex optimization problems that covers both finite-sum and expectation settings. Unlike several existing methods\, our algorithm has only a single loop without taking snapshots. It can achieve the best-known oracle complexity bounds using both constant and adaptive learning rates. We also discuss many variants of our algorithms. Then\, we demonstrate the flexibility of our estimators by applying them to approximate a stationary point of non-convex compositional optimization problems as well as a KKT point of minimax problems. In all cases\, our methods have a single loop and achieve a state-of-the-art oracle complexity. Finally\, we illustrate the proposed algorithms on several numerical examples using available and common datasets. \nThis talk is based on the collaboration with D. Liu (UNC)\, L. Nguyen (IBM)\, N. Pham (UNC)\, and D. Phan (IBM).
URL:https://stor.unc.edu/event/graduate-seminar-quoc-tran-dinh/
LOCATION:zoom
CATEGORIES:Graduate Seminar
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=America/New_York:20201005T160000
DTEND;TZID=America/New_York:20201005T171500
DTSTAMP:20220128T122708
CREATED:20200817T134044Z
LAST-MODIFIED:20200817T134044Z
UID:9942-1601913600-1601918100@stor.unc.edu
SUMMARY:STOR Colloquium: Patrick Combettes\, NCSU
DESCRIPTION:Patrick Louis Combettes\nNorth Carolina State University \n\nPerspective Functions and Applications \nIn this talk I will discuss mathematical and computational issues pertaining to perspective functions\, a powerful concept that permits to extend a convex function to a jointly convex one in terms of an additional scale variable. Applications in inverse problems and statistics will be presented.
URL:https://stor.unc.edu/event/stor-colloquium-patrick-combettes-ncsu/
LOCATION:Hanes Hall\, Hanes Hall\, Chapel Hill\, NC\, 27599\, United States
CATEGORIES:STOR Colloquium
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=America/New_York:20201019T160000
DTEND;TZID=America/New_York:20201019T171500
DTSTAMP:20220128T122708
CREATED:20200904T085346Z
LAST-MODIFIED:20200904T085346Z
UID:9944-1603123200-1603127700@stor.unc.edu
SUMMARY:STOR Colloquium: Lihua Lei\, Stanford
DESCRIPTION:Lihua Lei\nStanford University \n\nHierarchical Community Detection for Heterogeneous and \nMulti-scaled Networks \n \nReal-world networks are often hierarchical\, heterogeneous\, and multi-scaled\, while the idealized stochastic block models that are extensively studied in the literature tend to be over-simplified. In a line of work\, we propose several top-down recursive partitioning algorithms which start with the entire network and divide the nodes into two communities by certain spectral clustering methods repeatedly\, until a stopping rule indicates no further community structures. For these algorithms\, the number of communities does not need to be known a priori or estimated consistently. On a broad class of hierarchical network models motivated by Clauset\, Moore and Newman (2008)\, in which the communities are allowed to be heterogeneous and multi-scaled in terms of the size and link probabilities\, our algorithms are proved to achieve the exact recovery for sparse networks with expected node degrees logarithmic in the network size\, and are computationally more efficient than non-hierarchical spectral clustering algorithms. More interestingly\, we identify regimes where no algorithm can recover all communities simultaneously while our algorithm can still recover the mega-communities (unions of communities defined by the hierarchy) consistently without recovering the finest structure. Our theoretical results are based on my newly developed two-to-infinity eigenspace perturbation theory for binary random matrices with independent or dependent entries.
URL:https://stor.unc.edu/event/stor-colloquium-lihua-lei-stanford/
LOCATION:Hanes Hall\, Hanes Hall\, Chapel Hill\, NC\, 27599\, United States
CATEGORIES:STOR Colloquium
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=America/New_York:20201030T153000
DTEND;TZID=America/New_York:20201030T163000
DTSTAMP:20220128T122708
CREATED:20201026T131959Z
LAST-MODIFIED:20201026T131959Z
UID:9951-1604071800-1604075400@stor.unc.edu
SUMMARY:Grad Student Seminar: Adam Waterbury
DESCRIPTION:Stochastic Approximation of Quasi-Stationary Distributions \nWe propose two numerical schemes for approximating quasi-stationary distributions (QSD) of finite state Markov chains with absorbing states. Both schemes are described in terms of certain interacting chains in which the interaction is given in terms of the total time occupation measure of all particles in the system. The schemes can be viewed as combining the key features of the two basic simulation-based methods for approximating QSD originating from the works of Fleming and Viot (1979) and Aldous\, Flannery\, and Palacios (1998)\, respectively. In this talk I will describe the two schemes\, discuss their convergence properties as both time and the number of particles in the system simultaneously become large\, and present some exploratory numerical results comparing them to other QSD approximation methods.
URL:https://stor.unc.edu/event/grad-student-seminar-adam-waterbury/
LOCATION:zoom
CATEGORIES:Graduate Seminar
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=America/New_York:20201102T160000
DTEND;TZID=America/New_York:20201102T171500
DTSTAMP:20220128T122708
CREATED:20200904T085458Z
LAST-MODIFIED:20200904T085458Z
UID:9945-1604332800-1604337300@stor.unc.edu
SUMMARY:STOR Colloquium: Jacob Bien\, USC
DESCRIPTION:Jacob Bien\nUniversity of Southern California \n\nTree-Based Aggregation of Rare Features for Prediction \n \nIt is common in modern prediction problems for many features to be counts of rarely occurring events. The challenge posed by such “rare features” has received little attention despite its prevalence in diverse areas\, ranging from biology (e.g.\, rare species within a microbiome) to natural language processing (e.g.\, rare words within an online hotel review). We show\, both theoretically and empirically\, that not explicitly accounting for the rareness of features can greatly reduce the effectiveness of an analysis. We next propose a framework for aggregating rare features into denser features in a flexible manner that creates better predictors of the response. Applications to the microbiome and to online hotel reviews show how our methodology is useful in a wide range of contexts.
URL:https://stor.unc.edu/event/stor-colloquium-jacob-bein-usc/
LOCATION:Hanes Hall\, Hanes Hall\, Chapel Hill\, NC\, 27599\, United States
CATEGORIES:STOR Colloquium
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=America/New_York:20201106T153000
DTEND;TZID=America/New_York:20201106T163000
DTSTAMP:20220128T122708
CREATED:20201103T093422Z
LAST-MODIFIED:20201103T093422Z
UID:9952-1604676600-1604680200@stor.unc.edu
SUMMARY:Grad Student Seminar: Michael Conroy
DESCRIPTION:Michael Conroy\nUNC-Chapel Hill \nEfficient rare-event simulation for branching processes \nIn this talk I’ll discuss some of my past\, current\, and future work with importance sampling schemes for maxima of branching processes. In a recent paper\, my collaborators and I developed a strongly efficient and unbiased estimator for tail events of the maximum of a branching random walk with perturbation (or a Galton-Watson process on a random tree). The sampling procedure relies on a change of measure applied to the entire tree that randomly selects one branch to which it applies an exponential tilt\, leaving other branches unchanged. The process of selecting a single path suggests that alterations can be made to the estimator to reduce the computational complexity associated with a large branching rate. It also allows us to conjecture a conditional limit theorem that provides insight into how extreme events occur in branching random walks. I plan to make this talk very accessible\, starting with the basics of importance sampling and exponential tilting. This talk includes joint work with Mariana Olvera-Cravioto\, Bojan Basrak (University of Zagreb)\, and Zbigniew Palmowski (Wroclaw University of Science and Technology).
URL:https://stor.unc.edu/event/grad-student-seminar-michael-conroy-2/
LOCATION:zoom
CATEGORIES:Graduate Seminar
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=America/New_York:20201109T160000
DTEND;TZID=America/New_York:20201109T171500
DTSTAMP:20220128T122708
CREATED:20200904T085554Z
LAST-MODIFIED:20200904T085554Z
UID:9946-1604937600-1604942100@stor.unc.edu
SUMMARY:STOR Colloquium: Mayya Zhilova\, Georgia Tech
DESCRIPTION:Mayya Zhilova\nGeorgia Institute of Technology \n\nNonasymptotic Edgeworth-type expansions \nfor growing dimension. \n \nIn this talk I would like to discuss the problem of establishing higher order accuracy of bootstrapping procedures and (non-)normal approximation in the multivariate or high-dimensional setting. This topic is important for numerous problems in statistical inference and applications concerned with confidence estimation and hypothesis testing\, and involving a growing dimension of random data or unknown parameter. In particular\, I will focus on higher-order expansions for the uniform distance over the set of all Euclidean balls. The talk will include an overview of main ideas in the proofs\, and examples of statistical problems where the new results lead to improvements in accuracy of approximation. \nThe talk is based on the results in the recent preprint: https://arxiv.org/abs/2006.03959\, and https://arxiv.org/abs/1611.02686
URL:https://stor.unc.edu/event/stor-colloquium-mayya-zhilova-georgia-tech/
LOCATION:Hanes Hall\, Hanes Hall\, Chapel Hill\, NC\, 27599\, United States
CATEGORIES:STOR Colloquium
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=America/New_York:20201116T160000
DTEND;TZID=America/New_York:20201116T171500
DTSTAMP:20220128T122708
CREATED:20200817T134141Z
LAST-MODIFIED:20200817T134141Z
UID:9943-1605542400-1605546900@stor.unc.edu
SUMMARY:STOR Colloquium: Kavita Ramanan\, Brown University
DESCRIPTION:Kavita Ramanan\nBrown University \nLarge Deviations of Random Projections of \nHigh-dimensional Measures \n \nProperties of random projections of high-dimensional probability measures are of interest in a variety of fields\, including asymptotic convex geometry\, and high-dimensional statistics and data analysis. A particular question of interest is to identify what properties of the high-dimensional measure are captured by its lower-dimensional projections. While fluctuations of these projections have been well studied over the past decade\, we describe more recent work on both annealed and quenched large deviations principles for random projections\, and their refinements. \nThis talk is based on joint works with Nina Gantert\, Steven Soojin Kim and Yin-Ting Liao.
URL:https://stor.unc.edu/event/stor-colloquium-kavita-ramanan-brown-university/
LOCATION:Hanes Hall\, Hanes Hall\, Chapel Hill\, NC\, 27599\, United States
CATEGORIES:STOR Colloquium
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=America/New_York:20201120T153000
DTEND;TZID=America/New_York:20201120T163000
DTSTAMP:20220128T122708
CREATED:20201116T163043Z
LAST-MODIFIED:20201116T163043Z
UID:9953-1605886200-1605889800@stor.unc.edu
SUMMARY:Grad Student Seminar: Samopriya Basu\, Jack Prothero
DESCRIPTION:Samopriya Basu\nFiducial inference for SDEs with constant diffusion \nIn this talk\, I will talk about my research with my advisor Prof. Jan Hannig on carrying out fiducial inference for stochastic ordinary differential equations with constant diffusion coëfficient. The diffusion coëfficient σ is unknown and the drift term can depend on any number of unknown parameters β\, and the task is to come up with a data-dependent distribution estimator Fid(·) on the parameter space Θ ⊂ ℝ+ × ℝp+1\, called the fiducial distribution\, for the parameter vector θT = (σ\, βT) by “inverting” the law Pθ of the observable quantity of interest Q(θ)\, and to efficiently sample from it. I will discuss the challenges associated with carrying out fiducial inference in such settings and some of our resolutions thereof inspired by some approximation techniques in the Uncertainty Quantification literature. \nJack Prothero\nCentering in Functional Data Analysis \nData matrix centering is an ever-present yet underexamined aspect of data analysis. Centering such that data features have mean zero is often the default operation\, but in many contexts the standard centering practice is less clear and the broader consequences of centering are less understood. We present the effects of different forms of centering formally alongside a unified terminology for such operations. We then explore data where different forms of centering enhance interpretability of analytic results. Finally\, we present a novel statistical test to determine whether additional centering may be appropriate for a given data matrix.
URL:https://stor.unc.edu/event/grad-student-seminar-samopriya-basu-jack-prothero/
LOCATION:zoom
CATEGORIES:Graduate Seminar
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=America/New_York:20210129T120500
DTEND;TZID=America/New_York:20210129T130000
DTSTAMP:20220128T122708
CREATED:20210125T092825Z
LAST-MODIFIED:20210125T092825Z
UID:9955-1611921900-1611925200@stor.unc.edu
SUMMARY:Friday Lunch
DESCRIPTION:Informal lunch for the Fall 2020 graduate student cohort. Get to know each other better! Guests will include various faculty and graduate students from previous cohorts. Check your email for the Zoom link.
URL:https://stor.unc.edu/event/friday-lunch/2021-01-29/
LOCATION:Hanes Hall\, Hanes Hall\, Chapel Hill\, NC\, 27599\, United States
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=America/New_York:20210201T153000
DTEND;TZID=America/New_York:20210201T163000
DTSTAMP:20220128T122708
CREATED:20210125T092132Z
LAST-MODIFIED:20210125T092132Z
UID:9954-1612193400-1612197000@stor.unc.edu
SUMMARY:Colloquium: Dmitriy Drusvyatskiy\, University of Washington
DESCRIPTION:Dmitriy Drusvyatskiy\nUniversity of Washington at Seattle \nStochastic methods for nonsmooth nonconvex optimization \nStochastic iterative methods lie at the core of large-scale optimization and its modern applications to data science. Though such algorithms are routinely and successfully used in practice on highly irregular problems (e.g. deep neural networks)\, few performance guarantees are available outside of smooth or convex settings. In this talk\, I will describe a framework for designing and analyzing stochastic methods on a large class of nonsmooth and nonconvex problems\, with provable efficiency guarantees. The problem class subsumes such important tasks as phase retrieval\, robust PCA\, and minimization of risk measures\, while the methods include stochastic subgradient\, Gauss-Newton\, and proximal point iterations. The main thread of the proposed framework is appealingly intuitive. I will show that a wide variety of stochastic methods can be interpreted as inexact gradient descent on an implicit smoothing of the problem. Optimal learning rates and novel sample-complexity guarantees (for various signal and matrix recovery problems) follow quickly from this viewpoint. \nBisketch: Dmitriy Drusvyatskiy received his PhD from the Operations Research and Information Engineering department at Cornell University in 2013\, followed by a post doctoral appointment in the Combinatorics and Optimization department at Waterloo\, 2013-2014. He joined the Mathematics department at University of Washington as an Assistant Professor in 2014\, and was promoted to an Associate Professor in 2019. Dmitriy’s research broadly focuses on designing and analyzing algorithms for large-scale optimization problems\, primarily motivated by applications in data science. Dmitriy has received a number of awards\, including the Air Force Office of Scientific Research (AFOSR) Young Investigator Program (YIP) Award\, NSF CAREER\, INFORMS Optimization Society Young Researcher Prize 2019\, and finalist citations for the Tucker Prize 2015 and the Young Researcher Best Paper Prize at ICCOPT 2019. Dmitriy is currently a co-PI of the NSF funded Transdisciplinary Research in Principles of Data Science (TRIPODS) institute at University of Washington.
URL:https://stor.unc.edu/event/colloquium-dmitriy-drusvyatskiy-university-of-washington/
LOCATION:zoom
CATEGORIES:STOR Colloquium
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=America/New_York:20210205T120500
DTEND;TZID=America/New_York:20210205T130000
DTSTAMP:20220128T122708
CREATED:20210125T092825Z
LAST-MODIFIED:20210125T092825Z
UID:9956-1612526700-1612530000@stor.unc.edu
SUMMARY:Friday Lunch
DESCRIPTION:Informal lunch for the Fall 2020 graduate student cohort. Get to know each other better! Guests will include various faculty and graduate students from previous cohorts. Check your email for the Zoom link.
URL:https://stor.unc.edu/event/friday-lunch/2021-02-05/
LOCATION:Hanes Hall\, Hanes Hall\, Chapel Hill\, NC\, 27599\, United States
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=America/New_York:20210212T120500
DTEND;TZID=America/New_York:20210212T130000
DTSTAMP:20220128T122708
CREATED:20210125T092825Z
LAST-MODIFIED:20210125T092825Z
UID:9957-1613131500-1613134800@stor.unc.edu
SUMMARY:Friday Lunch
DESCRIPTION:Informal lunch for the Fall 2020 graduate student cohort. Get to know each other better! Guests will include various faculty and graduate students from previous cohorts. Check your email for the Zoom link.
URL:https://stor.unc.edu/event/friday-lunch/2021-02-12/
LOCATION:Hanes Hall\, Hanes Hall\, Chapel Hill\, NC\, 27599\, United States
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=America/New_York:20210217T153000
DTEND;TZID=America/New_York:20210217T163000
DTSTAMP:20220128T122708
CREATED:20210215T151942Z
LAST-MODIFIED:20210215T151942Z
UID:9976-1613575800-1613579400@stor.unc.edu
SUMMARY:Grad Student Seminar: Stefanos Kechagias\, SAS
DESCRIPTION:Stefanos Kechagias\nAnalytical Consulting & Enterprise Solutions\,\nSAS Institute \n\nScratch out “learn TensorFlow” from your New Year bucket list. In goes “learn how to tell valuable stories” \nIn this talk we will discuss three Story Telling Mediums statisticians employ to share their expert knowledge: scientific writing\, technical presentations and development of statistical software. First\, we will lay out proper foundations of scientific writing by placing the reader at the center of a statistician’s thought process. Under this reader-centric prism we will review examples of writing rules and discuss why we must know them\, how we should use them and when we can disregard them. We will then switch gears to technical presentations for audiences of mixed academic or industry backgrounds and for listeners and viewers of various technical competencies\, focusing on preparation and only briefly elaborating on delivery. We will conclude our ambitious threefold goal by shedding light on standard shortcomings of software development processes and practices that statisticians often adopt.
URL:https://stor.unc.edu/event/grad-student-seminar-stefanos-kechagias-sas/
LOCATION:zoom
CATEGORIES:Graduate Seminar
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=America/New_York:20210219T120500
DTEND;TZID=America/New_York:20210219T130000
DTSTAMP:20220128T122708
CREATED:20210125T092825Z
LAST-MODIFIED:20210125T092825Z
UID:9958-1613736300-1613739600@stor.unc.edu
SUMMARY:Friday Lunch
DESCRIPTION:Informal lunch for the Fall 2020 graduate student cohort. Get to know each other better! Guests will include various faculty and graduate students from previous cohorts. Check your email for the Zoom link.
URL:https://stor.unc.edu/event/friday-lunch/2021-02-19/
LOCATION:Hanes Hall\, Hanes Hall\, Chapel Hill\, NC\, 27599\, United States
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=America/New_York:20210305T120500
DTEND;TZID=America/New_York:20210305T130000
DTSTAMP:20220128T122708
CREATED:20210125T092825Z
LAST-MODIFIED:20210125T092825Z
UID:9960-1614945900-1614949200@stor.unc.edu
SUMMARY:Friday Lunch
DESCRIPTION:Informal lunch for the Fall 2020 graduate student cohort. Get to know each other better! Guests will include various faculty and graduate students from previous cohorts. Check your email for the Zoom link.
URL:https://stor.unc.edu/event/friday-lunch/2021-03-05/
LOCATION:Hanes Hall\, Hanes Hall\, Chapel Hill\, NC\, 27599\, United States
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=America/New_York:20210312T120500
DTEND;TZID=America/New_York:20210312T130000
DTSTAMP:20220128T122708
CREATED:20210125T092825Z
LAST-MODIFIED:20210125T092825Z
UID:9961-1615550700-1615554000@stor.unc.edu
SUMMARY:Friday Lunch
DESCRIPTION:Informal lunch for the Fall 2020 graduate student cohort. Get to know each other better! Guests will include various faculty and graduate students from previous cohorts. Check your email for the Zoom link.
URL:https://stor.unc.edu/event/friday-lunch/2021-03-12/
LOCATION:Hanes Hall\, Hanes Hall\, Chapel Hill\, NC\, 27599\, United States
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=America/New_York:20210315T153000
DTEND;TZID=America/New_York:20210315T163000
DTSTAMP:20220128T122708
CREATED:20210205T114105Z
LAST-MODIFIED:20210205T114105Z
UID:9970-1615822200-1615825800@stor.unc.edu
SUMMARY:STOR Colloquium: David Matteson\, Cornell
DESCRIPTION:Sparse Identification and Estimation of Large-Scale Vector AutoRegressive Moving Averages \nThe Vector AutoRegressive Moving Average (VARMA) model is fundamental to the theory of multivariate time series; however\, identifiability issues have led practitioners to abandon it in favor of the simpler but more restrictive Vector AutoRegressive (VAR) model. We narrow this gap with a new optimization-based approach to VARMA identification built upon the principle of parsimony. Among all equivalent data-generating models\, we use convex optimization to seek the parameterization that is simplest in a certain sense. A user-specified strongly convex penalty is used to measure model simplicity\, and that same penalty is then used to define an estimator that can be efficiently computed. We establish consistency of our estimators in a double-asymptotic regime. Our non-asymptotic error bound analysis accommodates both model specification and parameter estimation steps\, a feature that is crucial for studying large-scale VARMA algorithms. Our analysis also provides new results on penalized estimation of infinite-order VAR\, and elastic net regression under a singular covariance structure of regressors\, which may be of independent interest. We illustrate the advantage of our method over VAR alternatives on three real data examples. See https://arxiv.org/abs/1707.09208
URL:https://stor.unc.edu/event/stor-colloquium-david-matteson-cornell/
LOCATION:zoom
CATEGORIES:STOR Colloquium
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=America/New_York:20210317T153000
DTEND;TZID=America/New_York:20210317T163000
DTSTAMP:20220128T122708
CREATED:20210316T091454Z
LAST-MODIFIED:20210316T091454Z
UID:9977-1615995000-1615998600@stor.unc.edu
SUMMARY:Graduate Seminar: Alexander Murph
DESCRIPTION:Generalized Fiducial Inference on Differentiable Manifolds \nI’ll discuss the problem of defining a general fiducial density on an implicitly defined differentiable manifold and introduce our recent solution. Our proposed density extends the usual generalized fiducial distribution (GFD) by projecting the Jacobian differential onto the space that only allows directions of change that satisfy some constraint function. This calculation is shown to be simple to compute and exists under minor smoothness assumptions. To circumvent the need for an intractable marginal integral calculation\, we use two different constrained Monte Carlo algorithms that can efficiently explore a constrained parameter space. Then\, we consider several simple examples for which a direct parameterization exists and compare our density against these direct solutions. \nThis is a joint project with Jan Hannig and Jon Williams. In addition to discussing our recent work on this specific problem\, I’ll talk a bit about the general ideas of Hamiltonian Monte Carlo and Generalized Fiducial Inference.
URL:https://stor.unc.edu/event/graduate-seminar-alexander-murph/
LOCATION:zoom
CATEGORIES:Graduate Seminar
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=America/New_York:20210319T120500
DTEND;TZID=America/New_York:20210319T130000
DTSTAMP:20220128T122708
CREATED:20210125T092825Z
LAST-MODIFIED:20210125T092825Z
UID:9962-1616155500-1616158800@stor.unc.edu
SUMMARY:Friday Lunch
DESCRIPTION:Informal lunch for the Fall 2020 graduate student cohort. Get to know each other better! Guests will include various faculty and graduate students from previous cohorts. Check your email for the Zoom link.
URL:https://stor.unc.edu/event/friday-lunch/2021-03-19/
LOCATION:Hanes Hall\, Hanes Hall\, Chapel Hill\, NC\, 27599\, United States
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=America/New_York:20210322T153000
DTEND;TZID=America/New_York:20210322T163000
DTSTAMP:20220128T122708
CREATED:20210205T114157Z
LAST-MODIFIED:20210205T114157Z
UID:9971-1616427000-1616430600@stor.unc.edu
SUMMARY:STOR Colloquium: Linan Chen\, McGill
DESCRIPTION:A Glimpse into Random Geometry: from Brownian motion\nto Gaussian Free Field \nFor “random curve”\, a natural and classical model is Brownian motion; when it comes to “random surface”\, a promising candidate model is Gaussian free field (GFF)\, which can be seen as the analog of Brownian motion with multi-dimensional time parameters. In this talk\, we will introduce GFF from the viewpoint of infinite dimensional Gaussian measure\, discuss some problems arising from the study of geometric properties of GFF\, and explore the role of GFF in the construction of random geometry. \nThe field of random geometry (associated with GFF) is developing fast\, with many problems worth investigating. This talk could only give a (extremely limited) glimpse into this broad and rich topic.
URL:https://stor.unc.edu/event/stor-colloquium-linan-chen-mcgill/
LOCATION:zoom
CATEGORIES:STOR Colloquium
END:VEVENT
END:VCALENDAR