Skip to main content Skip to footer site map
Mathematics Department

Operations Research and Statistics Seminar

Fall 2023

  • Oct
    23
  • Marine Corps Prior Service Recruiting Optimization
    CAPT(s) Gary Lazzaro
    Time: 03:45 PM

    View Abstract

    We optimize the recruiting of prior service Marines into the Selected Marine Corps Reserve. Marine Corps Recruiting Command (MCRC) divides the US into six districts with 88 prior service recruiters (PSR) split between 20 offices to solicit prior service Marines. We solve two distinct problems. First, should the MCRC district boundary lines be changed to reflect current prior service recruiting challenges? Second, should locations of PSRs be changed in order to better contact prior service Marines? K-Means Clustering enables us to redraw the MCRC recruiting district lines to equitably distribute the recruit population. Our linear program optimizes the relocation of the PSRs to reserve center locations nationwide to minimize total distance traveled by recruiters. This research is one of the student projects for the 2 semester Naval Innovation Capstone course. This talk was originally given at the 2023 INFORMS annual meeting in Phoenix.
  • Oct
    11
  • Research Overview: Dan Bates and Chris Lourenco
    Dan Bates and Chris Lourenco
    USNA
    Location: CH 351
    Time: 03:45 PM

    View Abstract

    (Dan): Dan's past work mostly revolves around numerical methods for solving polynomial systems, sort of a step up from numerical linear algebra. He will give a birds-eye view of that part of his work, then say a bit about his interests in optimization and OR. In particular, he'll mention past work on optimal control (via KKT conditions) and some ideas for future work, possibly including integrating polynomial system solvers with existing OR techniques. (Chris): In this talk, Chris will give a brief overview of his work in exact linear programming, graph theory, and nonlinear optimization. Specifically, he has spent most of his time deriving a class of exact matrix factorization algorithms similar to those used inside of modern optimization solvers. He will discuss these methods as well as some recent work with USNA students on graph theory and nonlinear gas network optimization.
  • Oct
    05
  • The Future of Work Through the Lens of Machine Learning
    Chris Glynn
    Indeed
    Location: CH351
    Time: 03:45 PM

    View Abstract

    Recent advances in generative AI have raised urgent questions about the future of work. How will jobs change, and what skills will be required for jobseekers to be competitive? These questions are top of mind for individual jobseekers, employers, workforce development agencies, universities, and policy makers. In this talk, we utilize a family of Machine Learning models called Topic Models to investigate the evolution of skills advertised in job postings on Indeed. A family of Dynamic Linear Topic Models is developed to seamlessly unify traditional topic models for exchangeable data with Bayesian state-space models for temporally dependent data. We find that job postings are becoming both more technical while increasingly emphasizing the human aspects of work.
  • Sep
    11
  • Allocation of Tactical Search Assets in Undersea Warfare (and a bit of everything else!)
    Anna Svirsko
    US Naval Academy
    Location: CH351
    Time: 03:45 PM

    View Abstract

    In this talk, Anna will provide everyone with an introduction to some of her ongoing work and previous work. The talk will begin with current work done in collaboration with MIDN Sebastian Martin and Prof. Daphne Skipper on the allocation of tactical search assets in undersea warfare. It will briefly highlight the future work on integrating surveillance and tactical search and Bayesian search. Finally, she will provide an overview of other work she has done/is doing on concussions, disasters, and operational support airlift! protection resources to a set of facilities in order to maximize the expected coverage provided by the facilities. The probability that a facility fails depends on the amount of protection resources allocated to it. Exploiting the submodularity of the problem, we first develop a description of the hypograph of continuous submodular functions and then develop a cutting plane algorithm that finds approximate solutions and bounds. Next, we develop a spatial branch-and-bound approach that utilizes the approximate cutting plane algorithm to form an outer approximation and obtain upper bounds for a region. We present computational results that compare our method with a state-of-the-art solver.
go to Top