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:UTC
BEGIN:STANDARD
TZOFFSETFROM:+0000
TZOFFSETTO:+0000
TZNAME:UTC
DTSTART:20210101T000000
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTART;TZID=UTC:20211029T153000
DTEND;TZID=UTC:20211029T163000
DTSTAMP:20211028T205726
CREATED:20211027T152509Z
LAST-MODIFIED:20211027T152509Z
UID:12963-1635521400-1635525000@stor.unc.edu
SUMMARY:BIOS/STOR Student Seminar: Sam Rosin\, BIOS and Alex Murph\, STOR
DESCRIPTION:Sam Rosin\nPhD Student in BIOS\nAdvisor: Michael Hudgens \nEstimating Seroprevalence of SARS-CoV-2\nDuring the COVID-19 pandemic\, governments have used\nseroprevalence studies to estimate the proportion of persons\nwithin a given population who have antibodies to SARS-CoV-2.\nHowever\, serologic assays are prone to false positives and\nnegatives\, and non-probability sampling methods may induce\nselection bias. Our work considers seroprevalence estimators\nthat address both challenges by leveraging validation and\ncovariate data. We study the estimators in simulations and apply\nthem to SARS-CoV-2 seroprevalence studies in North Carolina\nand Belgium from 2020. \nAlexander Murph \nPhD Student in STOR\nAdvisors: Jan Hannig and\nJonathan P Williams \nBayesian Change Point Detection for Mixed Data\nwith Missing Values\nMany models in-production assume that incoming data follows the same\ndistribution of the data on which the model was trained. In practice\, however\,\nthis distribution will often not remain static. We develop a change-point\ndetection system on data over time to locate major changes in the underlying\ndata distribution. This system has a strong practical motivation\, since it can be\nused to warn data scientists that an in-production model’s performance may\ndrop: a direct consequence of the training data no longer representing the\ncurrent data. This can drive decisions such as when to retrain a model or when\nto investigate discrepancies in the data.\nWe will use a real-world example (the Mayo Clinic’s Palliative Care model) to\nmotivate the need for this system. Our solution takes a hierarchical Bayesian\napproach that learns a vector of regimes as an unknown parameter. These\nregime vectors are assumed to come from a Hidden Markov Model (HMM);\ngiven a regime vector\, data within a regime are modeled using random graphs\nwith a varying edge set. Missing values and mixed variable types are handled\nusing a Bayesian latent variable approach; parameters are learned using a\nvariation of the Split-Merge algorithm and a Double Reversible Jump\nMetropolis-Hastings algorithm.
URL:https://stor.unc.edu/event/bios-stor-student-seminar-sam-rosin-bios-and-alex-murph-stor/
LOCATION:133 Rosenau Hall\, Rosenau Hall\, Chapel Hill\, NC\, 27516\, United States
CATEGORIES:BIOS/STOR Student Seminar
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=UTC:20211101T080000
DTEND;TZID=UTC:20211101T100000
DTSTAMP:20211028T205726
CREATED:20211020T125737Z
LAST-MODIFIED:20211027T151849Z
UID:12956-1635753600-1635760800@stor.unc.edu
SUMMARY:Ph.D. Defense: Aman Barot
DESCRIPTION:Techniques in network embedding and Gaussian comparison for high-dimensional statistics \nThis dissertation consists of research on two high-dimensional statistical problems. In the first part of the dissertation\, we study Gaussian comparison which is an important technique for comparing distributions and functionals of Gaussian random variables. We derive a Gaussian comparison result based on a smart-path argument. We show the significance of this result by an application to a problem of maximal correlations in high dimensions. \n \nIn the second part of the dissertation\, we focus on the problem of community detection on networks using node embedding methods. In recent decades\, network data sets containing millions and billions of nodes have become available. This has necessitated the development of scalable methods for their analysis. One such class of methods are methods for node embedding. Node embedding methods encode nodes of a network in a low-dimensional Euclidean space which allows one to use well-known methods for Euclidean spaces for network analysis. In this dissertation we study the problem of community detection using two well-known node embedding methods: DeepWalk and node2vec. We describe the network sparsity regimes when the k-means algorithm applied to the node embeddings detects communities for graphs generated from the stochastic block model\, and when such an approach might fail. We also describe how increasing the non-backtracking parameter in the node2vec method leads to provable improvements in community detection.
URL:https://stor.unc.edu/event/ph-d-defense-aman-barot/
LOCATION:zoom
CATEGORIES:PhD Defense
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=UTC:20211101T153000
DTEND;TZID=UTC:20211101T163000
DTSTAMP:20211028T205726
CREATED:20210907T173640Z
LAST-MODIFIED:20211020T125304Z
UID:12808-1635780600-1635784200@stor.unc.edu
SUMMARY:STOR Colloquium: Thomas Lee\, University of California - Davis
DESCRIPTION:Thomas Lee\nUniversity of California\, Davis \nFiber Direction Estimation in Diffusion MRI \nDiffusion magnetic resonance imaging is a medical imaging technology to probe anatomical architectures of biological samples in an in vivo and non-invasive manner through measuring water diffusion. It is widely used to reconstruct white matter fiber tracts in brains. This can be done with the following steps. First\, one estimates the diffusion direction(s) for each voxel of the biological sample. As it is reasonable to assume that the diffusion directions from neighboring voxels are similar\, a local smoothing may be applied to the estimated tensors or directions to improve the estimation of diffusion directions. Finally\, a tracking algorithm is used to reconstruct fiber tracts based on (estimated) diffusion directions. \nMost commonly used tensor estimation methods assume a single tensor and do not work well when multiple principal diffusion directions are within a single voxel. The first part of this talk reports our work on identifying and estimating multiple diffusion directions within a voxel. The second part describes a direction smoothing method that greatly improves diffusion direction estimation in regions with crossing fibers. This smoothing method is shown to have excellent theoretical and empirical properties. Lastly\, this talk presents a fiber tracking algorithm that takes (estimated) diffusion directions as input and accommodates multiple directions within a voxel. \nThis is joint work with Seungyong Hwang\, Debashis Paul\, Jie Peng\, and Raymond Wong.
URL:https://stor.unc.edu/event/stor-colloquium-thomas-lee-university-of-california-davis/
CATEGORIES:STOR Colloquium
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=UTC:20211104T100000
DTEND;TZID=UTC:20211104T120000
DTSTAMP:20211028T205726
CREATED:20211020T125528Z
LAST-MODIFIED:20211028T152927Z
UID:12951-1636020000-1636027200@stor.unc.edu
SUMMARY:Ph.D. Defense: Kevin O'Connor
DESCRIPTION:Kevin O’Connor\nComputation and Consistent Estimation of Stationary Optimal Transport Plans \n In this dissertation\, we study optimal transport (OT) for stationary stochastic processes\, a field that we refer to as stationary optimal transport. Through example and theory\, we argue that when applying OT to stationary processes\, one should incorporate the stationarity into the problem directly\, constraining the set of allowed transport plans to those that are stationary themselves. In this way\, we only consider transport plans that respect the dependence structure of the marginal processes. We study this constrained OT problem from statistical and computational perspectives\, with an eye toward applications in machine learning and data science. In particular\, we \n\ndevelop algorithms for computing stationary OT plans of Markov chains.\nextend these tools for Markov OT to the alignment and comparison of weighted graphs.\npropose estimates of stationary OT plans based on finite sequences of observations.\n\nWe build upon existing techniques in OT as well as draw from a variety of fields including Markov decision processes\, graph theory\, and ergodic theory. In doing this\, we uncover new perspectives on OT and pave the way for additional applications and approaches in future work. \n
URL:https://stor.unc.edu/event/phd-defense-kevin-oconnor/
LOCATION:405 Dey Hall
CATEGORIES:PhD Defense
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=UTC:20211108T153000
DTEND;TZID=UTC:20211108T163000
DTSTAMP:20211028T205726
CREATED:20210824T165600Z
LAST-MODIFIED:20211019T171135Z
UID:12767-1636385400-1636389000@stor.unc.edu
SUMMARY:STOR Colloquium: Lauren Davis\, North Carolina A&T University
DESCRIPTION:Title: Improving equity and access in hunger relief supply chains: models for prediction and distribution of uncertain supply \nAbstract: \nDuring the past decade\, an increasing number of natural disasters and humanitarian emergencies have prompted significant research in the area of relief chain logistics and supply chain management. Much of the research has focused on challenges associated with stocking and distribution of relief supplies in response to sudden-onset disasters. However\, issues surrounding slow onset and persistent disasters (like food insecurity) present a unique set of challenges\, particularly with respect to the management and distribution of donated supply. Based on a partnership with a local non-profit hunger relief organization\, we describe the relief supply chain associated with the provision of food aid to populations suffering from hunger. We present predictive and descriptive models that quantify the availability of supply over time\, characterize demand\, and optimize the distribution of uncertain supply to ensure equity and improve access. Implications of our findings on operational efficiency and service delivery are discussed. \n \nBio: \nDr. Lauren Davis is a professor in the Department of Industrial & Systems Engineering at North Carolina Agricultural and Technical State University. She received her B.S. in Computational Mathematics from Rochester Institute of Technology; M.S. in Industrial Engineering from Rensselaer Polytechnic Institute\, and Ph.D. in Industrial Engineering from NC State University. Her research focuses on decision-making under uncertainty primarily using stochastic optimization techniques (Markov Decision Processes\, stochastic programming) and simulation. Her work has been applied to solve optimal stocking\, transportation scheduling and distribution decisions in for-profit and non-profit supply chains. She has more than 40 peer-reviewed journal papers and refereed conference proceedings addressing issues related to inventory management\, transportation scheduling\, port operations\, and emergency response in areas such as food supply chains\, food security\, port operations\, and humanitarian relief. Her work has been supported by the National Science Foundation\, Department of Homeland Security\, and US Department of Agriculture totaling more than $4 million in grant funding. Additionally\, her research examining hunger relief supply chains has been featured in CNN’s Great Big Story and NSF’s Discovery article series. She is currently the Principal Investigator for an NSF-funded National Research Traineeship grant that explores food security and hunger relief using computational data science.
URL:https://stor.unc.edu/event/stor-colloquium-lauren-davis-nc-at/
LOCATION:120 Hanes Hall\, Hanes Hall\, Chapel Hill\, NC\, 27599\, United States
CATEGORIES:STOR Colloquium
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=UTC:20211110T090000
DTEND;TZID=UTC:20211110T110000
DTSTAMP:20211028T205726
CREATED:20211020T125637Z
LAST-MODIFIED:20211020T125637Z
UID:12954-1636534800-1636542000@stor.unc.edu
SUMMARY:Ph.D. Defense: Samopriya Basu
DESCRIPTION:
URL:https://stor.unc.edu/event/ph-d-defense-samopriya-basu/
CATEGORIES:PhD Defense
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=UTC:20211115T133000
DTEND;TZID=UTC:20211115T153000
DTSTAMP:20211028T205726
CREATED:20210917T173709Z
LAST-MODIFIED:20210917T173709Z
UID:12837-1636983000-1636990200@stor.unc.edu
SUMMARY:Ph.D. Defense: Nhan Huu Pham
DESCRIPTION:
URL:https://stor.unc.edu/event/ph-d-defense-nhan-huu-pham/
LOCATION:130 Hanes Hall\, Hanes Hall\, Chapel Hill\, 27599\, United States
CATEGORIES:PhD Defense
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=UTC:20211122T153000
DTEND;TZID=UTC:20211122T163000
DTSTAMP:20211028T205726
CREATED:20210824T165649Z
LAST-MODIFIED:20211018T151159Z
UID:12769-1637595000-1637598600@stor.unc.edu
SUMMARY:STOR Colloquium: George Lan\, Georgia Institute of Technology
DESCRIPTION:Stochastic optimization methods for reinforcement learning \nABSTRACT: Reinforcement Learning (RL) has attracted considerable interest from both industry and academia recently. The study of RL algorithms with provable rates of convergence\, however\, is still in its infancy. In this talk\, we discuss some recent progresses on the solutions of two fundamental RL problems\, i.e.\, stochastic policy evaluation and policy optimization\, based on our studies for stochastic optimization methods. More specifically\, we develop novel analysis of temporal difference (TD) learning and present new conditional TD algorithm (CTD) and fast TD (FTD) algorithm to achieve the best-known so-far convergence rate for policy evaluation. For policy optimization\, we introduce a new class of policy mirror descent (PMD) methods and show that they achieve linear convergence for the deterministic case and optimal sampling complexity for the stochastic case\, regardless whether the RL problem is regularized or not. \nBIO: Guanghui (George) Lan is an A. Russell Chandler III professor in the H. Milton Stewart School of Industrial and Systems Engineering at Georgia Institute of Technology. Dr. Lan was on the faculty of the Department of Industrial and Systems Engineering at the University of Florida from 2009 to 2015\, after earning his Ph.D. degree from Georgia Institute of Technology in August 2009. His main research interests lie in optimization and machine learning. The academic honors he received include the Mathematical Optimization Society Tucker Prize Finalist (2012)\, INFORMS Junior Faculty Interest Group Paper Competition First Place (2012) and the National Science Foundation CAREER Award (2013). Dr. Lan serves as an associate editor for Mathematical Programming\, SIAM Journal on Optimization and Computational Optimization and Applications. He is also an associate director of the Center for Machine Learning at Georgia Tech.
URL:https://stor.unc.edu/event/stor-colloquium-george-lan-georgia-tech/
LOCATION:120 Hanes Hall\, Hanes Hall\, Chapel Hill\, NC\, 27599\, United States
CATEGORIES:STOR Colloquium
END:VEVENT
END:VCALENDAR