CS Seminars

Abstract:

Today, we are seeing an explosion in emerging compute/data-intensive applications, such as AI and data analysis services, the rollout of next-generation networks, and the growth of smart edge devices. However, the transition to a post-Moore era raises significant concerns ranging from compute and storage efficiency to carbon footprint/energy consumption. An underexplored but promising opportunity to improve the cost-performance-sustainability tradeoffs of existing computing systems is the use of approximation. Data systems may not need to calculate results with 100% precision to maintain operational reliability.

In this talk, I will present my research on scaling computing systems with approximation techniques for various analytical tasks across the computing stack, such as network traffic analysis and dynamic connected data processing. First, I will describe how bridging theory and practice with sketching and sampling techniques can significantly speed up network analytics under tight resource budgets. Second, I will discuss efficient algorithms and system optimizations that enable mining complex structures in large-scale graph data. The systems I have built are backed with rigorous theoretical guarantees and achieve several orders of magnitude improvements with small accuracy losses. The developed network analytics solution is the first of its kind deployed in popular open-source network processing libraries (e.g., Data Plane Development Kit), and the approximate graph systems are under evaluation in the industry (e.g., a fintech company and a cloud provider). Finally, I will chart paths to designing future approximate computing systems with heterogeneous hardware that balance performance, reliability, and sustainability.

Biography:

Alan Zaoxing Liu is an Assistant Professor in Electrical and Computer Engineering at Boston University. His work spans computer systems, networks, and applied algorithms to co-design performant, reliable, and secure data analytics solutions across the computing stack. His recent research focuses on designing scalable and trustworthy approximate computing systems. He is a recipient of the best paper award at FAST'19 and received interdisciplinary recognitions, including ACM STOC "Best-of-Theory" plenary talk and USENIX ATC "Best-of-Rest". Previously, he did postdoctoral research at Carnegie Mellon University CyLab and received his Ph.D. in Computer Science from Johns Hopkins University.

//

Abstract:

A long-standing vision in computer science has been to evolve computing devices into proactive assistants that enhance our productivity, health and wellness, and many other facets of our lives. User digitization is crucial in achieving this vision as it allows computers to intimately understand their users, capturing activity, pose, routine, and behavior. Today’s consumer devices – like smartphones and smartwatches – provide a glimpse of this potential, offering coarse digital representations of users with metrics such as step count, heart rate, and a handful of human activities like running and biking. Even these very low-dimensional representations are already bringing value to millions of people’s lives, but there is significant potential for improvement. In my research, I develop new algorithms and methods that allow consumer devices to capture rich, continuous representations of their users. Armed with such knowledge, our future devices could offer longitudinal health tracking, more productive work environments, full-body avatars in extended reality, and embodied telepresence experiences, to name just a few domains. Critically, these advances cannot come at the expense of user practicality, meaning my work must be strategic in developing new sensors and making use of existing sensors and edge computation.

Biography:

Karan is a Ph.D. candidate at the School of Computer Science at Carnegie Mellon University, specializing in novel sensing and interaction techniques. In his thesis work, Karan focused on increasing the fidelity of user digitization technologies while retaining or improving user practicality, opening new paradigms in augmented and virtual reality, health monitoring, natural user interfaces, and context-aware computing. Many of his research projects have been open-sourced, deployed in-the-wild, licensed by tech companies, and even shipped as a product feature. To date, Karan has published over 25 papers at top venues. He is a Siebel Fellow and the Editor-in-Chief of ACM Crossroads (XRDS). His research has been widely covered in the media, including NBC Nightly News, Today Show, CNN, TechCrunch, Engadget, NPR, Fast Company, and Gizmodo among others.

CS Department Events

CS Graduate Student Appreciation Brunch

Tech Talk Series: Agile Methodology: A Glimpse into the Tech World's Development Tool

IDEAL Workshop on Machine Learning, Interpretability, and Logic

Other Events & Opportunities

McCormick Summer Undergraduate Research Awards

All students awarded a McCormick Summer Research Award will be given a stipend of $4500, intended to defray summer living costs. It is expected that students will devote ~8 weeks of full-time effort to the project. If necessary for the project, students may apply for additional funds (up to $500) to cover research-related expenses. In this case, a budget should be provided explaining how the additional funds are to be used.

To submit your proposal, use this link. The deadline for submission is 5:00pm, April 3, 2023.

(Note that McCormick's summer research program is distinct from similar opportunities offered by Northwestern's Office of Undergraduate Research.)

CASMI Virtual Panel: "The Harms and Benefits of Generative AI: Exploring the Differences Between Fluency and Fact"

CASMI AI@NU Research Series: Daniel Martin - Analyzing Machine Learning using Cognitive Economic Methods

Thought Leader Dialogue on AI Education

Artificial intelligence (AI) is becoming increasingly integrated across our technology ecosystem, yet public understanding of AI is limited. AI education initiatives are needed to aid people in making informed decisions about their personal technology use, AI-related policy decisions, and the role AI should play in their schools and workplaces. In addition, widespread AI literacy has the potential to equip the public with skills needed to collaborate and create with rapidly changing AI technologies. Join Northwestern’s Center for Human-Computer Interaction + Design  for an engaging conversation about fostering public AI literacy through initiatives such as incorporating AI in K-12 education, designing novel and creative AI learning experiences, and engaging adults in learning and decision-making about AI in their communities. This panel will explore best practices for designing AI learning experiences, how to ensure access to AI education is inclusive and equitable, and what AI education may look like across different populations. Each speaker will give a short presentation on their research related to AI education.

Speakers include Dr. Cynthia Breazeal (MIT), Dr. Christina Gardner-McCune (University of Florida), and Dr. Ken Holstein (Carnegie Mellon). Presentations will be followed by a panel discussion moderated by Dr. Duri Long, Northwestern’s Center for Human Computer Interaction + Design.

WildHacks Mentorship Application

Data Science Challenges in Gravitational Wave Surveys

Gravitational wave (GW) astronomy has transformed from a niche area of speculative research to an invaluable component of the astronomer's toolkit. As access to the GW spectrum broadens, so too must the computational approaches used to extract information from the data. The Laser Interferometer Space Antenna (LISA) is an ESA-led mission to survey the mHz regime of the GW spectrum richly populated with a variety of galactic, extragalactic, and cosmological sources. LISA will also be a multimessenger power house, with copious sources expected to produce counterparts covering the electromagnetic spectrum. Fully realizing the science capabilities of LISA demands new data science approaches for GW signal extraction to cope with the numerous overlapping sources simultaneously present in the data. These include high- and trans-dimensional optimization algorithms for Bayesian model selection, machine learning algorithms for compactly representing multidimensional probability distributions, and hardware acceleration with GPUs to improve computational and energy efficiency for the pipelines. This talk will summarize the LISA mission, its science objectives, and take an in-depth look at the nuances, computational challenges, and current state of the art for processing and interpreting the LISA data.

News