News & Events
Department Events

Events

  • May
    27

    IEMS Nemhauser & Nelson Award Seminar

    Department of Industrial Engineering and Management Sciences (IEMS)

    10:45 AM A230, Technological Institute

    EVENT DETAILS

    Paper Title: An algorithm for stochastic convex-concave fractional programs with applications to production efficiency and equitable resource allocation

    Shibshankar Dey

    Co-authors: Cheolmin Kim, Sanjay Mehrotra

    Citation: This paper considers a relatively less studied optimization modeling framework of fractional optimization that is natural for modeling productivity and equity as a ratio of output to input. While deterministic problems are themselves challenging, the stochastic counterparts of these problems were intractable if global optimality is desired. With the idea of branching on a variable introduced for each fractional term this paper moves the needle on these very hard problems. The results in this paper demonstrate that reasonable solutions with provable bounds can be obtained within a few hours of computation, whereas commercial solvers fail to solve these problems.

    Paper Title: Improving self-training under distribution shifts via anchored confidence with theoretical guarantees

    Taejong Joo

    Co-author: Diego Klabjan

    Citation: This paper makes significant contributions in self-training with distribution shift. The latter happens frequently in practice (the environment or data change in time requiring making adjustments to a model). The idea of this paper's approach is to iteratively refine soft labels used in training. The strategy is based on previous labels and new statistical estimates of labels. One of the importances of this work is the fact that modeling and algorithmic choices are backed by mathematical analyses. The algorithm in this paper consistently outperforms existing algorithms.

    Paper Title: Sample-Path Large Deviations for Lévy Processes and Random Walks with Lognormal Increments

    Zhe Su

    Co-author: Chang-Han Rhee

    Citation: The theory of large deviations has a long and successful history of providing systematic tools for analyzing rare events. However, the classical large deviations framework often falls short for heavy-tailed stochastic systems. The three most widely used classes of distributions for modeling heavy tails are regular variation, heavy-tailed Weibull, and lognormal. Over the past decade, the regularly varying and heavy-tailed Weibull cases have been successfully addressed for Lévy processes and random walks. However, the lognormal case has remained open until recently. This paper resolves this gap by establishing large deviations for Lévy processes and random walks with lognormal increments. Zhe’s subsequent projects demonstrate the strength of these results by applying them to successfully analyze queue-length asymptotics in two classical queueing systems: stochastic fluid networks and many-server queues.

    more

    TIME Tuesday, May 27, 2025 at 10:45 AM - 12:00 PM

    LOCATION A230, Technological Institute    map it

    ADD TO CALENDAR

    CONTACT Kendall Minta    kendall.minta@gmail.com EMAIL

    CALENDAR Department of Industrial Engineering and Management Sciences (IEMS)