Hanes Hall

Alfredo Garcia, Army Research Office (ARO) and University of Florida   Iterative Mechanisms for Electricity Markets   Abstract: We consider the problem of designing the rules by which dispatch and payment to electricity market participants are gradually adjusted while taking into account network and reliability constraints so as to ensure the market clears with an efficient outcome. Small adjustments (which require minimal information from market participants at each iteration) facilitate the identification of incentives for ensuring truthful reporting of private…

Bin Hu University of North Carolina at Chapel Hill, Kenan-Flagler Business School Is Reshoring Better than Offshoring? The Effect of Offshore Supply Dependence In this paper we investigate the effect of offshore supply dependence (OSD) on offshoring-reshoring profit comparisons. We find that OSD hampers a reshoring manufacturer's responsiveness to demand information updates and may significantly affect offshoring-reshoring comparisons, such that reshoring may yield lower profits than offshoring in many cases, including when offshoring has no baseline-cost advantage. We then show…

Nelson Antunes CEMAT/University of Lisbon and University of Algarve Sampling Internet traffic: estimation of flow distributions In this talk, the focus is on recovering the flow (sequence of packets sharing common attributes) size and duration distributions from sampled Internet traffic using a number of previously introduced packet sampling algorithms. Assuming a basic probabilistic flow model, the distribution of flow durations is expressed in terms of the distributions of flow sizes and flow IATs (interarrival times between packets). The available estimation…

Joe Guinness North Carolina State University Permutation methods for sharpening Gaussian process approximations   Vecchia's approximate likelihood for Gaussian process parameters depends on how the observations are ordered, which can be viewed as a deficiency because the exact likelihood is permutation-invariant. I take the alternative standpoint that the ordering of the observations is an aspect that can be tuned to sharpen the approximations. I show that advantageously chosen orderings of the observations can drastically improve the approximations. In addition to…

Dirk Lorenz Braunschweig Technical University Probabilistic Image Models and Extensions of the Perona-Malik Filter The Perona-Malik model has been very successful at restoring images from noisy input.  In this paper, we show how the Perona-Malik model can be reinterpreted and extended using the language of Gaussian scale mixtures.  Specifically, we show how the expectation-maximization EM algorithm applied to Gaussian scale mixtures leads to the lagged-diffusivity algorithm for computing stationary points of the Perona-Malik diffusion equations.  Moreover, we show how mean…

Impact of Delay Announcements in Call Centers: An Empirical Approach   We undertake an empirical study of the impact of delay announcements on callers' abandonment behavior and the performance of a call center with two priority classes. A Cox regression analysis reveals that in this call center, callers' abandonment behavior is affected by the announcement messages heard. To account for this, we formulate a structural estimation model of callers' (endogenous) abandonment decisions. In this model, callers are forward-looking utility maximizers…

Ivo Adan Technical University of Eindhoven and EURANDOM The Netherlands   A rate balance principle   We introduce a rate balance principle for general (not necessarily Markovian) stochastic processes, with special attention to birth-death like processes. This principle appears to be useful in deriving well-known, as well as new results for various queueing systems.           Refreshments will be served at 3:00pm in the 3rd floor lounge of Hanes Hall

This colloquium has been cancelled.  We apologize for any inconvenience. STOR colloquium: Vanja Dukic, University of Colorado-Boulder

Stephen Becker University of Colorado Boulder Efficient robust PCA algorithms for the GPU   We introduce the matrix completion problem and the similar robust PCA (RPCA) problem and discuss their relation to compressed sensing and some of their applications to collaborative filtering, background detection in videos, and neuroscience. We cover some standard algorithms to solve these problems, including proximal gradient descent, Frank-Wolfe/conditional-gradient, and Burer-Monteiro splitting. A natural idea to speed up the algorithms is to run them on the GPU…

Title: Commuting, voting, and the off-label use of data   Abstract: The modern practice of data science often involves approaching datasets differently than in traditional statistics.  Two analyses will be presented, both of which use combinations of public datasets in expected and "off-label" ways.  One addresses the need to simulate realistic commuting patterns, and the other evaluates the effects of North Carolina voting site decisions.  In the process of telling the stories of these two data science projects, we'll have…

Santiago Balseiro Duke University Dynamic Mechanisms with Martingale Utilities We study the dynamic mechanism design problem of a seller who repeatedly sells independent items to a buyer with private values. In this setting, the seller could potentially extract the entire buyer surplus by running efficient auctions and charging an upfront participation fee at the beginning of the horizon. In some markets, such as internet advertising, participation fees are not practical since buyers expect to inspect items before purchasing them. This…

Julie Ivy North Carolina State University To be Healthy, Wealthy, and Wise: Using Decision Modeling to Personalize Policy in Health, Hunger Relief, and Education   Decision making to satisfy the basic human needs of health, food, and education is complex, accomplishing this in such a way that solutions are personalized to the needs of the individual has been the focus of my research. This talk will present an overview of my research in the area of decision making under conditions…

Peter Hoff Duke University Adaptive FAB confidence intervals with constant coverage   Abstract: Confidence intervals for the means of multiple normal populations are often based on a hierarchical normal model. While commonly used interval procedures based on such a model have the nominal coverage rate on average across a population of groups, their actual coverage rate for a given group will be above or below the nominal rate, depending on the value of the group mean.   In this talk…

Gudrun Johnsen University of Iceland   Title: If you look close enough you can see a whole lot: Data collection and analysis of the Parliamentary Investigation Commission looking into the Icelandic Banking Collapse in 2008   Abstract: The combined collapse of Iceland's three largest banks is the third largest bankruptcy in history and the largest banking system collapse suffered by any country in modern economic history, relative to GDP. In 2008 the Icelandic Parliament installed a Special Investigation Commission (SIC)…

Directed complex networks and ranking algorithms In the first part of this talk I will discuss a family of inhomogeneous directed random graphs for modeling complex networks such as the web graph, Twitter, ResearchGate, and other social networks. This class of graphs includes as a special case the classical Erdos-Renyi model, and can be used to replicate almost any type of predetermined degree distributions, in particular, power-law degrees such as those observed in most real-world networks. I will mention during…

Zhiyi Zhang University of North Carolina at Charlotte Title: Statistical Implications of Turing’s Formula   Abstract: This talk is organized into three parts.   Turing’s formula is introduced. Given an iid sample from a countable alphabet under a probability distribution, Turing’s formula (introduced by Good (1953), hence also known as the Good-Turing formula) is a mind-bending non-parametric estimator of total probability associated with letters of the alphabet that are NOT represented in the sample. Many of its statistical properties were…

Hongtu Zhu University of North Carolina at Chapel Hill, and The University of Texas MD Anderson Cancer Center   Title: Statistical Challenges, Opportunities, and Strategies in Large-Scale Medical Studies With the rapid growth of modern technology, many biomedical studies have collected data across different sources (e.g., imaging, genetics, and clinical) in an unprecedented scale. The integration of such ultra high-dimensional data raises many statistical challenges, rendering most existing statistical methods and old data platform no longer suitable and thus underscoring…

Philip Ernst Rice University   Title: Yule's "Nonsense Correlation" Solved!   Abstract: In this talk, I will discuss how I recently resolved a longstanding open statistical problem. The problem, formulated by the British statistician Udny Yule in 1926, is to mathematically prove Yule's 1926 empirical finding of ``nonsense correlation.”  We solve the problem by analytically determining the second moment of the empirical correlation coefficient of two independent Wiener processes. Using tools from Fredholm integral equation theory, we calculate the second…

Ilse Ipsen North Carolina State University Randomized Algorithms for Matrix Computations   The emergence of massive data sets, over the past fifteen or so years, has led to the development of Randomized Numerical Linear Algebra. Fast and accurate randomized matrix algorithms are being designed for applications like machine learning, population genomics, astronomy, nuclear engineering, and optimal experimental design. We give a flavour of randomized algorithms for the solution of least squares/regression problems and, if time permits, for the computation of…

Title: Different Paradigms of Interpretation for Forensic Value of Evidence Quantification Abstract: Currently, one of the major problems in the forensic science community is the confusion between different statistical paradigms. A quantification of the value of evidence is interpreted differently under each paradigm, and may even be the answer to different questions. It is our opinion that these issues need to be addressed before quantitative forensic analyses are considered a reliable science in the justice system. A related issue is the…

Jonathan M. Lees University of North Carolina at Chapel Hill   Geophysical Time Series Analysis on Volcanoes: Can we quantify non-linearity? Most geophysical processes are aperiodic noisy, intermittent and transient.  This requires specialized methods for time series analysis, that seek patterns in time series that vary in space and time.  I present here examples from research on exploding volcanoes that exhibit tremor that appears to be resonant but likely results from nonlinear feedback systems.  The physical models for these observations…

Output-Weighted Active Sampling for Bayesian Uncertainty Quantification and Prediction of Rare Events Themis Sapsis We 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…

Patrick Louis Combettes North Carolina State University Perspective Functions and Applications In 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.

Lihua Lei Stanford University Hierarchical Community Detection for Heterogeneous and Multi-scaled Networks   Real-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…

Jacob Bien University of Southern California Tree-Based Aggregation of Rare Features for Prediction   It 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…

Mayya Zhilova Georgia Institute of Technology Nonasymptotic Edgeworth-type expansions for growing dimension.   In 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…

Kavita Ramanan Brown University Large Deviations of Random Projections of High-dimensional Measures   Properties 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…