Opening

Opening Remarks

Conference
8:30 AM — 9:00 AM CST
Local
Jul 26 Mon, 8:30 PM — 9:00 PM EDT

Opening Remarks

Ness Shroff, Huadong Ma, Xin Liu, Edmund Yeh

0
This talk does not have an abstract.

Session Chair

Huadong Ma (BUPT)

Keynote-1

Keynote: Wireless Networking with Deadlines

Conference
9:00 AM — 10:00 AM CST
Local
Jul 26 Mon, 9:00 PM — 10:00 PM EDT

Keynote: Wireless Networking with Deadlines

P. R. Kumar (TAMU)

0
Electromagnetic waves undergo multiple uncontrollable alterations as they propagate within a wireless environment. Free space path loss, signal absorption, as well as reflections, refractions, and We consider the problem of operating wireless networks when packets have deadlines. We present a characterization of capacity as well as optimal scheduling policies for the cases of an access point as well as a network of point-to-point links operating over unreliable channels.

Session Chair

P. R. Kumar (TAMU)

Keynote-2

Keynote: Applications and Security of Vehicular Communications

Conference
10:00 AM — 11:00 AM CST
Local
Jul 26 Mon, 10:00 PM — 11:00 PM EDT

Keynote: Applications and Security of Vehicular Communications

Kang Shin (Umich)

0
As we are entering the era of autonomous driving, vehicular communication remains to be one of the most important development focuses for both governments and industries because it is the backbone for establishing future Intelligent Transportation Systems (ITSs). This talk consists of two main parts. In the first part, I will present the widely-perceived application scenarios of vehicular communications, including vehicle-to-vehicle (V2V), vehicle-to-infrastructure (V2I), and vehicle-to-everything (V2X), and how they can enhance the safety, mobility and eco-friendliness of future transportation systems. I will also discuss the security threats in vehicular communications and the potential solutions to those threats. In the second part, I will address the acute problem of data verification in vehicular communications. Because using incorrect/compromised information in ITSs can degrade traffic throughput or potentially cause fatal accidents, it is crucial that the receiving end of vehicular communications, especially the infrastructure, must ensure the correctness of the received information. I will discuss the important properties and functions that should be considered in designing a system for data verification. Next, I will introduce DeBi, a novel way of verifying the integrity of vehicular data. Specifically, DeBi determines if the vehicle state data (e.g., vehicle location, speed, acceleration, etc.) carried in the V2I/V2X messages is trustworthy for use in ITSs. It is designed to meet the various requirements of different ITSs and facilitate their future deployment while reflecting app constraints/regulations and adapting to changing environments. It features Adaptive Detection Scope, Adaptive Run-time Operation, Adaptive Deployment, and Individual Data Verification & Reconstruction. Finally, I will conclude this talk by discussing the unresolved issues and potential research directions in vehicular communications.

Session Chair

Kang Shin (Umich)

Keynote-3

Keynote: So Many Choices, So Little Time: Foundations and Applications of Comparison and Ranking Models

Conference
1:45 PM — 2:45 PM CST
Local
Jul 27 Tue, 1:45 AM — 2:45 AM EDT

Keynote: So Many Choices, So Little Time: Foundations and Applications of Comparison and Ranking Models

Matthias Grossglauser (EPFL)

0
The digital world offers infinite possibilities, which forces us to make choices all the time: we click on a search result in a list, choose a song from iTunes, or pick a restaurant on Yelp. Because of the central role of choosing, there is an abundance of data capturing comparisons and rankings. This has led to a resurgence of interest in discrete-choice models in the machine learning community. In this talk, we discuss several recent results in this context.

We discuss large-scale inference of the Plackett-Luce (PL) model, a widely used probabilistic choice model, via an iterative spectral algorithm. We also consider the PL model in an active learning setting, where we can ask an oracle comparison questions and receive noisy answers. Our goal is to recover the underlying ranking accurately by asking as few questions as possible. When we constrain choices to a network setting, capturing the process of navigating on a graph using local information, we obtain a new node metric termed ChoiceRank that estimates link traffic (or strength) more faithfully than other alternatives, such as PageRank.

We then explore several variations of choice models embedded in concrete applications. In peer production systems, such as a large collaborative software project or a parliament writing laws, such models combine predictive performance with interpretable parameters. In interactive search, we are able to find a target in a large database without formulating a query.

Session Chair

Matthias Grossglauser (EPFL)

Made with in Toronto · Privacy Policy · MobiHoc 2020 · © 2021 Duetone Corp.