Time: 12-1pm on Fridays

Location: EBII 3211

11/22/2024 — Invited Talk

Presenter: Prof. Huining Li

Title: Harnessing Mobile Technologies for Healthcare Equity and Fairness

Abstract:

Mobile health technologies are increasingly recognized as a means to bridge
health disparities due to their high accessibility, cost-effectiveness, and global
connectivity. However, research indicates that if these technologies are not
implemented thoughtfully, they could exacerbate health disparities and raise issues,
such as privacy and fairness. In my talk, I will discuss my recent efforts in addressing
these challenges. Firstly, I will present a novel mobile health framework designed to
protect privacy while efficiently managing and assessing neural disorders, such as
Parkinson's disease. Secondly, I will introduce innovative, fairness-aware computational
methods for mobile mental health applications. Both of these applications have been
successfully deployed in real-world settings, actively involving participants in delivering
equitable and responsible mobile health services.

Bio:

Dr. Huining Li is an Assistant Professor of Department of Computer Science in the North
Carolina State University. She received Ph.D. degree from the Department of Computer
Science and Engineering, University at Buffalo in 2024. Her research interest lies
broadly in internet-of-things, cybersecurity, and mobile computing. Her work has
received three Best Paper Awards (SenSys’19, BodyNet’21, and ICHI’22), and one Best
Paper Candidate (SenSys’22). Huining was named among the EECS Rising Stars in
2023, and received the Harold O. Wolf Achievement Award in 2024.

11/15/2024 — Cancelled due to Haven’s schedule conflict

Presenter: Haven Brown (from Justin Bradley’s group)

Title:

Abstract:

Bio:

11/08/2024

Presenter: Abdullah Al Arafat (from Dr. Zhishan Guo's group) [Video]

Room: EBII 3001 (changed)

Title: Towards Secure and Resilient Real-Time Intelligent Systems

Abstract:

Recent advances in sensing, communication, and computing have significantly enhanced the accessibility and integration of Cyber-Physical Systems (CPS), which are often constrained by stringent timing requirements and scalability challenges. This talk presents my Ph.D. research on enhancing CPS resilience and security. To improve temporal resilience, I developed real-time scheduling methods using a mixed-criticality system model and created dynamic priority schedulers for the Robot Operating System (ROS 2). On the security front, I designed backdoor-resilient algorithms to enable secure use of pretrained machine learning models in resource-constrained CPS, addressing vulnerabilities associated with both in-domain and different than the inference domain pretrained models.

Bio:

Abdullah is a Computer Science Ph.D. candidate at NC State University, advised by Dr. Zhishan Guo. His research interests include real-time systems, robust learning, and cyber-physical systems.

11/01/2024 — Invited Talk (Virtual) [Slides] [Video]

CS Department announcement page: [HERE].

Time: 12-1pm EST

Location: EBII 3211

Presenter: Ed Younis @ Lawrence Berkeley National Lab

Title: Compiling Resource Efficient Programs with BQSKit

Abstract:

Quantum hardware is experiencing a boon leading to more chip variety and configurations with higher fidelities. While ultimately, this will translate to a boon for the entire field of quantum computing, it presents a software design problem by placing more of the overall burden of realizing end-to-end quantum applications on the software stacks, specifically the quantum compiler. The Berkeley Quantum Synthesis Toolkit (BQSKit) is a powerful and portable quantum compiler framework with a proven ability to alleviate this issue and translate recent hardware successes up to the algorithm level. BQSKit achieves superior portability and optimization potential by utilizing a parameterized quantum circuit intermediate representation to facilitate numerical instantiation. In this talk, I first introduce the idea of numerical instantiation and BQSKit compilation, including algorithms and workflows for transpiling circuits to any hardware, even ones with heterogeneous gate sets or higher-level qudits (qutrits). I then detail several further practical use cases, such as error mitigation techniques with approximations and algorithm-hardware design exploration.

Bio:

Ed is a computer systems engineer at Lawrence Berkeley National Laboratory with extensive experience developing and implementing advanced algorithms for quantum compilation, such as QFAST and Qfactor. He is currently the principal engineer on the BQSKit project and has research interests in quantum synthesis, compilation, and software systems.

10/25/2024 — Invited Talk (In-Person) [Slides]

CS Department announcement page: [HERE].

Time: 12-1pm on Fridays

Location: EBII 3211

Presenter: Grey Ballard @ Wake Forest University

Title: Randomized Algorithms for Tensor Decompositions

Abstract:

Tensor decompositions are generalizations of low-rank matrix approximations to higher dimensional data. They have become popular for their utility across applications—including blind source separation, dimensionality reduction, compression, anomaly detection—where the original data is represented as a multidimensional array. We’ll highlight a few applications where tensor decompositions, such as CP, Tucker, and Tensor Train decompositions, are particularly effective. We’ll discuss properties of the various decompositions, and we’ll describe the algorithms used to compute them. In particular, we’ll see how to apply randomized algorithms to reduce computational costs of the algorithms with minimal degradation of accuracy.

Bio:

Grey Ballard is an Associate Professor in the Computer Science Department at Wake Forest University. After receiving his PhD in computer science from the University of California Berkeley in 2013, he was a Truman Fellow at Sandia National Laboratories in Livermore, CA. He received his BS in math and computer science at Wake Forest in 2006 and his MA in math at Wake Forest in 2008.
His research interests include numerical linear algebra, high performance computing, and computational science, particularly in developing algorithmic ideas that translate to improved implementations and more efficient software. His work has been recognized with the Wake Forest Excellence in Research Award; an NSF CAREER award; the SIAM Linear Algebra Prize; three conference best paper awards, at SPAA, IPDPS, and ICDM; the C.V. Ramamoorthy Distinguished Research Award at UC Berkeley; and the ACM Doctoral Dissertation Award - Honorable Mention.

10/18/2024 (Cancelled)

10/11/2024 — Invited Talk (In-Person) [Video] [Slides]

CS Department announcement page: [HERE].

Time: 12-1pm on Oct 11

Location: EBII 3211

Presenter: Nikolaos Sidiropoulos @ University of Virginia

Title: Canonical correlation analysis through the lens of linear and multilinear algebra

Abstract:

Canonical correlation analysis (CCA) is a classic tool in statistics and machine learning. It is used to mine for pairs of highly correlated random variables that are mixed in two datasets. CCA is widely used and arguably the most important example of multiview analysis. At the same time, CCA and its generalizations exhibit remarkable richness and depth - including fundamental open questions - when examined from different points of view, which are often missed. For example, under what conditions will CCA recover a random variable that is linearly mixed in all the views? In this talk, we will explore the fundamental concepts and principles of CCA, and discuss practical applications including wireless communications (spectrum reuse / underlay; cell-edge detection) and fMRI (multi-subject common task-related spatio-temporal response recovery). We will focus on an algebraic interpretation of CCA, through a natural generative model that links CCA to coupled matrix factorization and sheds new light on CCA identifiability and algorithms. Easy to check identifiability conditions will be discussed, along with an algebraic interpretation of CCA as (approximate) subspace intersection. Time permitting, we will also discuss nonlinear (``deep'') CCA from a multilinear algebra / probabilistic graphical model point of view.

Short Bio:

Nicholas D. Sidiropoulos (Fellow, IEEE) received the Diploma in electrical engineering from the Aristotle University of Thessaloniki, Thessaloniki, Greece, and the M.S. and Ph.D. degrees in electrical engineering from the University of Maryland at College Park, College Park, MD, USA, in 1988, 1990, and 1992, respectively. He is the Louis T. Rader Professor with the Department of ECE, University of Virginia. He has previously served as a Faculty with the University of Minnesota and the Technical University of Crete, Greece. His research interests are in signal processing, communications, optimization, tensor decomposition, and machine learning. He received the NSF/CAREER award in 1998, IEEE Signal Processing Society (SPS) Best Paper Award for 2001, 2007, 2011, and 2022, and the IEEE SPS Donald G. Fink Overview Paper Award for 2022. He served as a IEEE SPS Distinguished Lecturer (2008–2009), the Vice President—Membership of the IEEE SPS (2017–2019), and the Chair of the SPS Fellow Evaluation Committee (2020–2021). He received the 2010 IEEE SPS Meritorious Service Award, the 2013 Distinguished Alumni Award of the ECE Department, University of Maryland, the 2022 EURASIP Technical Achievement Award, and the 2022 IEEE SPS Claude Shannon–Harry Nyquist Technical Achievement Award. He is a fellow of EURASIP (2014).

10/08/2024 — Invited Talk (In-Person) [Video]

CS Department announcement page: [HERE].

Presenter: Ramakrishnan Kannan @ Oak Ridge National Laboratory

Time: 3-4pm

Location: EBII 3211

Title: Knowledge-guided Machine Learning: Bridging Scientific Knowledge and AI

Abstract:

Scientific knowledge-guided machine learning (KGML) is an emerging field of research where scientific knowledge is deeply integrated in ML frameworks to produce solutions that are scientifically grounded, explainable, and likely to generalize on out-of-distribution samples even with limited training data. In this talk, we will demonstrate using both scientific knowledge and data as complementary sources of introduction in the design, training, and evaluation of ML models on scientific domains that is of interest to Department of Energy and ORNL. Particularly, we will show some of our recent research in the area of material and battery design.

Bio:

Dr. Ramakrishnan (Ramki) Kannan is a distinguished scientist leading the Discrete Algorithms group at Oak Ridge National Laboratory (ORNL). His research expertise spans distributed machine learning and graph algorithms on High-Performance Computing (HPC) platforms, focusing on accelerating scientific discovery by significantly reducing computation times, often from weeks to seconds.

Dr. Kannan's notable achievements include leading the DSNAPSHOT project for COVID-19, a finalist for the ACM Gordon Bell Award in 2020 and 2022, Summit ranking 3rd on the Graph500 benchmark and achieved 1 ExaFLOPS on a KnowledgeGraph AI application on Frontier and  UT-Battelle Research Accomplishment Award in 2023.

With a track record of securing over $8M in research funding and leading projects exceeding $1 million for the Department of Defense, Dr. Kannan currently serves as the Deputy Director for the DOE Mathematical Multifaceted Integrated Capability Center (MMICC) [Sparsitute](https://sparsitute.lbl.gov/). He co-authored "Knowledge-guided Machine Learning" with Prof. Anuj Karpatne of Virginia Tech and Prof. Vipin Kumar of the University of Minnesota, a significant publication in 2022. Dr. Kannan holds over 24 patents issued by the USPTO and has been recognized as an IBM Master Inventor. He earned his Ph.D. under Professor Haesun Park at Georgia Institute of Technology and his M.Sc (Engg) under Professor Y. Narahari at the Indian Institute of Science.

09/27/2024 — Invited Talk (In-Person) — [Video] [Slides]

CS Department announcement page: [HERE].

Zoom Link for NCSU Attendance:

Presenter: Yifan Sun @ William & Mary

Title: Towards Building Explainable Computer Architecture

Abstract:
Research in computer architecture has predominantly emphasized technical innovation, focusing on speed, execution time, power consumption, security, and reliability. While this technical focus has driven advancements in the field, it has also led to increasingly complex chip designs, creating a bottleneck for further innovation. One critical aspect often overlooked is explainability—the ability to comprehend mechanisms without excessive analysis. This gap in understanding is a primary root cause for various issues in hardware designs, including correctness bugs, performance bottlenecks, security vulnerabilities, reliability weak points, and sustainability concerns. We believe that rather than solely relying on technical advancement, the community needs to consider the relationship between computers and the architects who design them. It is essential to understand how computer architects work and design developer-friendly tools to facilitate their design. In this talk, Dr. Sun will discuss his research on building developer-friendly tools for computer architects, including the Akita simulator engine, AkitaRTM—the real-time monitoring tool for Akita, and Daisen—the visualization tool for Akita.

Short Bio:
Dr. Yifan Sun is an Assistant Professor in the Department of Computer Science at William & Mary. Dr. Sun received my Ph.D. degree in Computer Engineering from Northeastern University and an MS degree in Electrical Engineering from the University at Buffalo. His research spans computer architecture, data visualization, and human-computer interaction.

09/20/2024 — [Video] [Slides]

Presenter: Kurt Wilson (from Dr. Zhishan Guo's group)

Title: Physics-Aware Mixed-Criticality Systems Design via End-to-End Verification of CPS

Abstract: Autonomous systems are heavily used in many safety-critical systems, such as industrial automation, autonomous cars, Industrial Internet of Things (I-IoT), etc. Verification of the functional and temporal correctness of such systems is necessary before deployment to ensure their safety. However, due to the presence of physical systems in the continuous-time domain and computational models in the discrete-time domain, end-to-end verification of these systems is highly challenging. Existing formal methods focus on verifying physical models assuming static or simplified computation models. In contrast, existing real-time systems focus on satisfying strict timing bounds but do not care how those bounds are obtained and how they relate to physical safety. Our approach bridges these two domains,  and constitutes an end-to-end verification framework for arbitrary physical models and computational models incorporated within a cyber-physical automated system. By allowing the interaction between the computational and physical models, our verification framework enables a fine-grained scheme that verifies against the local environment instead of verifying against global worst-case assumptions. To support locally varying worst-case scenarios, a mixed-criticality system is proposed where the system supports several critical models and switches among the modes based on environmental uncertainty.

Bio: Kurt Wilson is a PhD student in the Real Time Intelligent Systems lab, advised by Dr. Zhishan Guo. Kurt works with real time systems, reinforcement learning, and formal verification.

09/13/2024 — [Video]

Presenter: Abdullah Al Arafat (from Dr. Zhishan Guo's group)

Title: Dynamic Priority Scheduling of Multi-Threaded ROS 2 Executor with Shared Resources (accepted at EMSOFT 2024)

Abstract: The second generation of Robot Operating System (ROS 2) received significant attention from the real-time system research community, mostly aiming at providing formal modeling and timing analysis. However, most of the current efforts are limited to the default scheduling design schemes of ROS 2. The unique scheduling policies maintained by default ROS 2 significantly affect the response time and acceptance rate of workload schedulability. It also invalidates the adaptation of the rich existing results related to non-preemptive (and limited-preemptive) scheduling problems in the real-time systems community to ROS 2 schedulability analysis. This paper aims to design, implement, and analyze a standard dynamic priority-based real-time scheduler for ROS 2 while handling shared resources. Specifically, we propose to replace the readySet with a readyQueue, which is much more efficient and comes with improvements for callback selection, queue updating, and a skipping scheme to avoid priority inversion from resource sharing. Such a novel ROS 2 executor design can also be used for efficient implementations of fixed-priority policies and mixed-policy schedulers. Our modified executor maintains the compatibility with default ROS 2 architecture. We further identified and built a link between the scheduling of limited-preemption points tasks via the global earliest deadline first (GEDF) algorithm and ROS 2 processing chain scheduling without shared resources. Based on this, we formally capture the worst-case blocking time and thereby develop a response time analysis for ROS 2 processing chains with shared resources. We evaluate our scheduler by implementing our modified scheduler that accepts scheduling parameters from the system designer in ROS 2. We ran two case studies–one using real ROS 2 nodes to drive a small ground vehicle, and one using synthetic tasks. The second case study identifies a case where the modified executor prevents priority inversion. We also test our analysis with randomly generated workloads. In our tests, our modified scheduler performed better than the ROS 2 default.

Bio: Abdullah is a Computer Science Ph.D. candidate at NC State University, advised by Dr. Zhishan Guo. His research interests include real-time systems, robust learning, and cyber-physical systems.

09/06/2024 — [Video] [Slides] [Paper]

Presenter: Blake Burgstahler (from Dr. Frank Mueller's group)

Title:
Synthesis of Approximate Parametric Circuits for Variational Quantum Algorithms

Abstract:
This work presents a novel approach to synthesize approximate circuits for the ansatze of variational quantum algorithms (VQA) and demonstrates its effectiveness in the context of solving integer linear programming (ILP) problems. Synthesis is generalized to produce parametric circuits in close approximation of the original circuit and to do so offline. This removes synthesis from the (online) critical path between repeated quantum circuit executions of VQA. We hypothesize that this approach will yield novel high fidelity results beyond those discovered by the baseline without synthesis. Simulation and real device experiments complement the baseline in finding correct results in many cases where the baseline fails to find any and do so with on average 32% fewer CNOTs in circuits.

Bio:
Blake Burgstahler is a fourth year PhD student in the Operations Research program at North Carolina State University (NCSU). His research focuses on the applications and algorithm development for quantum computing, focusing specifically on improving the efficacy of quantum computing for solving classical optimization problems via Variational Quantum Algorithms.

Prior to beginning his research work, Blake was a teaching assistant and instructor of record for various undergraduate mathematics courses, including Calculus, Linear Algebra, and Operations Research.

Blake holds a Master of Science in Applied Mathematics from the University of Washington, where he focused on data driven analysis methods and was a teaching assistant for Calculus courses. He received a Bachelor of Science in (Applied) Physics and a Bachelor of Arts in Mathematics from Bethel University in St. Paul, MN.

Upcoming Talks:

11/29/2024 — Cancelled Due to Thanksgiving Holiday