Keynotes and Tutorials


Sandjai Bhulai - Vrije Universiteit Amsterdam

Title: Harnessing machine learning for solving complex mathematical optimization problems

Abstract: We often encounter intricate problems in mathematical optimization that defy traditional solution approaches. These optimization challenges find applications across diverse sectors of society, ranging from logistics and transportation to healthcare and finance. Despite their ubiquity, the computational complexity of these problems can make them formidable to tackle.

In this presentation, we delve into the intersection of mathematical optimization and machine learning, demonstrating how the synergy between these two fields can lead to innovative solutions with tangible real-world applications. We will explore a spectrum of optimization problems that have proven notoriously difficult to solve using classical methods. We will showcase how machine learning techniques can be leveraged to enhance optimization processes, making them more efficient, accurate, and adaptable.

Nick Feamster - University of Chicago

Title: Fifteen Years of Measuring Access Network Performance: From Benchmarks to Equity

Abstract: The last 15 years have seen significant advances in the area of measuring broadband access networks. In the mid-2000s, measuring access networks was simpler: access networks were typically the throughput bottleneck along an end-to-end path, and access speeds were slower, making it far simpler to both measure ISP performance and to understand the contributions of the access ISP to overall end-to-end performance. Today, as access network speeds have increased, performance bottlenecks may lie anywhere along an end-to-end path, making performance analysis more challenging. In addition, higher  access throughput has moved attention to other aspects of performance, including latency and application performance, both of which are more challenging to measure. Finally, the increasing importance of the Internet in everyday life has led to a growing interest in the equity of Internet performance, leading to new challenges in measuring (and comparing) the performance across different geographies. In this talk, I will discuss this history of measuring access network performance, offering a retrospective on past challenges, a current overview of the state of the art in access network measurement, and a view into where the field is heading.


Giulia Fanti - Carnegie Mellon University

Title: The Theory and Practice of (Private) Synthetic Data

Abstract: Organizations are often unable to share data due to regulatory, business, and privacy concerns. The resulting data silos seriously inhibit the development, tuning, testing, and auditing of data science pipelines. In this talk, I will discuss the promise and challenges of using synthetic data from deep generative models to share data across institutional boundaries. This talk will begin by introducing several leading generative models for synthetic data. We will then explore how these generative models address (or do not address) some of the key requirements of synthetic data. In particular, we will focus on the privacy properties of synthetic data, and state-of-the-art techniques for generating differentially-private synthetic data.

Siva Theja Maguluri, Sushil Varma - Georgia Institute of Technology

Title: Stochastic Matching Networks: Theory and Applications

Abstract: The theory of stochastic processing networks and queues has evolved into a mature field over the last century with numerous applications. Unlike a classical queue, where a server stays put and serves customers, matching queues involve the arrivals and departures of both customers and servers. Some of the most basic questions about the stability of such systems turn out to be nontrivial. While the theory of matching queues and Stochastic Matching Networks is still in its early stages, there has been a lot of development in the last decade. Such an increased interest is due to their widespread applications in emerging domains like the gig economy, online marketplaces, electric vehicles, quantum switches, assemble-to-order systems, and payment channel networks. This tutorial will provide an overview of the developments in this space and expand on a few results and applications in more detail.

The tutorial will start with a short introduction and present a heavy-traffic theory of a single matching queue. This theory will be used to study pricing and matching algorithms in a ride-hailing system that involves bipartite matchings. After that, the setting of multipartite matchings will be presented along with an application in Quantum Switches. Finally, the tutorial will conclude with an overview of various related models and results on matching networks.

Online user: 1 Privacy