Women in AI

The AI-EDGE Institute conducts a number of activities to increase the participation of women in the field of AI, including a series of presentations by women researchers on a range of AI topics.  A summary of past presentations is shown below.

March 28,2024@11:00am EST

Towards Smart Health Using Mobile AI

Yingying (Jennifer) Chen
Department of Electrical and Computer Engineering
Rutgers University

Abstract

Smart health is crucial in promoting individual well-being and long-term health outcomes. With the increasingly intelligent sensing capabilities and emerging mobile AI in pervasive mobile devices, smart health technologies can empower individuals to monitor their health metrics, track their physical activity, and make informed decisions about their lifestyle, forming a powerful synergy that fosters healthier lifestyles and prevents chronic illnesses in the future. We find that commercial WiFi signals can be exploited as a sensing modality in addition to its original communication usage, providing a contactless and low-cost solution for smart health in non-clinical environments. This talk will first introduce a personalized fitness assistant system that utilizes WiFi signals and applied machine learning for effective exercise monitoring and assessment at a relatively coarse-grained level. The system leverages Channel State Information (CSI) measurements from existing WiFi infrastructure and AI techniques to provide workout statistics and dynamic evaluations. A Deep Neural Network (DNN) model is employed for workout recognition and individual identification tasks. The study investigates the impact of factors such as the sensitive region between WiFi transceivers and ambient interference on system performance. The talk will then take a deeper look of estimating fine-grained vital signs (e.g., breathing rate and heartbeats) during sleep using minute WiFi signal changes. Our approach demonstrates the feasibility of contactless, continuous, and fine-grained monitoring of vital signs without additional cost. In addition, the system can distinguish different sleep events and track sleep postures to provide insights into sleep quality. We show that vital signs can be captured using only one AP and a single WiFi device, which can be extended to non-sleep scenarios. Extensive experiments in laboratory and nonclinical settings show comparable performance compared to existing clinical approaches. These AI-based smart health systems can offer convenience and potential for various smart health application scenarios, benefiting users in maintaining their healthy daily routines.

Biography

Yingying (Jennifer) Chen is a Professor and Department Chair of Electrical and Computer Engineering (ECE) and Peter Cherasia Endowed Faculty Scholar at Rutgers University. She is the Associate Director of Wireless Information Network Laboratory (WINLAB). She also leads the Data Analysis and Information Security (DAISY) Lab. She is a Fellow of National Academy of Inventors (NAI), a Fellow of ACM and a Fellow of IEEE. She is also an ACM Distinguished Scientist. Her research interests include Applied Machine Learning in Mobile Computing and Sensing, Internet of Things (IoT), Security in AI/ML Systems, Smart Healthcare, and Deep Learning on Mobile Systems. She is a pioneer in RF/WiFi sensing, location systems, and mobile security. Before joining Rutgers, she was a tenured professor at Stevens Institute of Technology and had extensive industry experiences at Nokia (previously Lucent Technologies) before joining academia. She has published 3 books, 4 book chapters and 300+ journal articles and refereed conference papers. She is the recipient of seven Best Paper Awards in top ACM and IEEE conferences. She is the recipient of NSF CAREER Award and Google Faculty Research Award. She received NJ Inventors Hall of Fame Innovator Award and is also the recipient of IEEE Region 1 Technological Innovation in Academic Award. Her research has been supported by many funding agencies including NSF, NIH, ARO, DoD and AFRL and reported in numerous media outlets including MIT Technology Review, CNN, Fox News Channel, Wall Street Journal, National Public Radio and IEEE Spectrum. She has been serving/served on the editorial boards of IEEE Transactions on Mobile Computing (TMC), IEEE Transactions on Wireless Communications (TWireless), IEEE/ACM Transactions on Networking (ToN) and ACM Transactions on Privacy and Security (TOPS). For more information, please refer to her homepage at: www.winlab.rutgers.edu/~yychen/.

March 18, 2024@1:00pm EST

Revisiting Standards to Allow Innovation

Muriel Médard
Electrical Engineering and Computer Science Department
Massachusetts Institute of Technology

Abstract

What is the role of standards in future communications? In this talk, we present a vision of standards to help ensure both reliability and room for innovation. We argue that standards can successfully concentrate on purely functional matters, relying on modular APIs, rather than being prescriptive about methods, which often embed highly inefficient legacy technologies at a time when such waste is no longer tenable. We provide concrete illustrations of the possibilities of this construct, based on reliable transport with network coding and on universal decoding chips.

Biography

Muriel Médard is the NEC Professor of Software Science and Engineering in the Electrical Engineering and Computer Science (EECS) Department at MIT, where she leads the Network Coding and Reliable Communications Group in the Research Laboratory for Electronics at MIT and Chief Scientist for Steinwurf, which she has co-founded. She obtained three Bachelor's degrees, as well as her M.S. and Sc.D, all from MIT. Muriel is a Member of the US National Academy of Engineering (elected 2020), a Member of the German National Academy of Sciences Leopoldina (elected 2022), a Fellow of the US National Academy of Inventors (elected 2018), American Academy of Arts and Sciences (elected 2021), and a Fellow of the Institute of Electrical and Electronics Engineers (elected 2008). She holds Honorary Doctorates from the Technical University of Munich (2020) and from The University of Aalborg (2022).

Muriel was awarded the 2022 IEEE Kobayashi Computers and Communications Award. She received the 2019 Best Paper award for IEEE Transactions on Network Science and Engineering, the 2018 ACM SIGCOMM Test of Time Paper Award, as well as nine conference paper awards. 

Muriel currently serves as the Editor-in-Chief of the IEEE Transactions on Information Theory. Muriel was elected president of the IEEE Information Theory Society in 2012, and serves on its board of governors, having previously served for dozens years.

Muriel received the inaugural 2013 MIT EECS Graduate Student Association Mentor Award, voted by the students. She also received the inaugural MIT Postdoctoral Association (PDA) Mentor Award, noted by all current and post postdocs of MIT. She set up the Women in the Information Theory Society (WithITS) and Information Theory Society Mentoring Program, for which she was recognized with the 2017 Aaron Wyner Distinguished Service Award.

Muriel has over eighty US and international patents awarded, the vast majority of which have been licensed or acquired. For technology transfer, she has co-founded CodeOn, for which she consults, and Steinwurf, for which she is Chief Scientist.

Muriel has supervised over 40 master students, over 20 doctoral students and over 25 postdoctoral fellows.

March 8, 2024@3:00pm EST

Optimizing Action Space Grouping for Sample and Computation Efficient RL

Yining Li
Department of Electrical and Computer Engineering
The Ohio State University

Abstract

In this talk, we delve into the challenging realm of reinforcement learning (RL) in high-dimensional spaces, a domain often hindered by the curse of dimensionality. Our work introduces a novel approach that partitions action spaces into multiple groups based on similarities in transition distribution and reward function, coupled with a linear decomposition model to manage intra-group differences effectively. This method exploits the inherent structure of groupwise-similar  Markov Decision Processes (MDPs). While refined grouping strategies decrease approximation error, they concurrently increase estimation error under sample size constraints and computational resources. We propose an optimization problem to identify an optimal grouping strategy that balances performance loss with sample/computational complexity, offering a computationally efficient method for selecting a nearly optimal grouping strategy, irrespective of the action space size. Our theoretical analysis and experimental results underscore the significance of grouping strategies as a crucial degree of freedom in optimizing RL performance, paving the way for more effective and efficient RL applications in complex environments.

Biography

Yining Li is a Ph.D. student in the Department of Electrical and Computer Engineering at The Ohio State University, where she is advised by Prof. Ness Shroff. Her primary research focus is on reinforcement learning and non-convex optimization.

Panel Discussion on “Emerging AI Topics”

May 19, 2023@2:00pm EDT

We are excited to host four distinguished panelists for our upcoming event at AI-EDGE institute from the "Women in AI and Networking" group. We will be having an intriguing conversation among the panelists, hosts, and the audience about different trendy and emerging AI tools which have raised concerns and controversy recently.

The Panelists are:

1. Ness Shroff (OSU) : shroff.11@osu.edu

2. Sanjay Shakkottai (UT Austin): sanjay.shakkottai@utexas.edu

3. Sujata Banerjee (VmWare): sujatab@vmware.com

4. Nageen Himayat (Intel Labs): nageen.himayat@intel.com

Revisiting Attention Weights as Explanations from an Information Theoretic Perspective

February 15, 2023@3:00pm EST

K.P. (Suba) Subbalakshmi
Department of Electrical and Computer Engineering
Stevens Institute of Technology

Abstract

Over the years, machine learning (ML) and artificial intelligence (AI) models have steadily grown in complexity, accuracy, and other quality metrics, often at the expense of interpretability of the results. Simultaneously, researchers and practitioners have begun to realize that more transparency in the deep learning and artificial intelligence engines are necessary if the power of these engines should be adopted in practice. For example, having a very good performance metric for a disease predictor is of little use, if it is not possible to give an explanation to the end user (a physician, the patient or even the designer of the tool) on why the model came to the conclusions that it did. Attention mechanisms have recently gained popularity as a relatively less complex, strong model and attention scores have been used as a mechanism for explaining the model decisions. However, there is a debate on whether attention weights can, in fact, be used to identify the most important inputs to a model. We approach this question from an information theoretic perspective by measuring the mutual information between the model output and the hidden states. In this talk we will ask the following questions: (1) can attention mechanisms be used for model explanation? (2) does complexity of the attention mechanism affect explainability? (3) can attention mechanisms be taught to assign higher weights to the inputs that contain more information? And (4) do encoder choices affect explainability?

Biography

K.P. (Suba) Subbalakshmi is a Professor of Electrical and Engineering at Stevens Institute of Technology. She a Fellow of the National Academy of Inventors, a Jefferson Science Fellow, and a Member of the National Academy of Science Engineering and Medicines' Intelligence Science and Technology Experts Group. She is a recipient of the New Jersey Inventors Hall of Fame, Innovator Award.

Her current research interests include Artificial Intelligence and Machine Learning with an emphasis on explainable AI, mental health, cyber safety/security, and edge computing. She serves as an Associate Editor of the IEEE Transactions on Artificial Intelligence and the IEEE Transactions on Neural Networks and Learning Systems.

The Good and The Ugly: How Robots and AI Influence Human Behavior

Monday, January 30, 2023 at 1:00PM EST

Ayanna Howard
Dean of Engineering
The Ohio State University
Monte Ahuja Endowed Dean's Chair

Abstract

People tend to overtrust sophisticated computing devices, especially those powered by AI. As these systems become more fully interactive with humans during the performance of day-to-day activities, ethical considerations in deploying these systems must be more carefully investigated. Bias, for example, has often been encoded in and can manifest itself through AI algorithms, which humans then take guidance from, resulting in the phenomenon of excessive trust. Bias further impacts this potential risk for trust, or overtrust, in that these intelligent systems are learning by mimicking our own thinking processes, inheriting our own implicit gender and racial biases, for example. These types of human-machine feedback loops may consequently have a direct impact on the overall quality of the interaction between humans and machines, whether the interaction is in the domains of healthcare, job-placement, or other high-impact life scenarios. In this talk, we will discuss various forms of human overtrust with respect to these intelligent machines and possible ways to mitigate the impact of bias in our interactions with them.

Biography

Dr. Ayanna Howard is the Dean of Engineering at The Ohio State University and Monte Ahuja Endowed Dean's Chair. Previously she was the Linda J. and Mark C. Smith Endowed Chair in Bioengineering and Chair of the School of Interactive Computing at the Georgia Institute of Technology. Dr. Howard’s research encompasses advancements in artificial intelligence (AI), assistive technologies, and robotics, and has resulted in over 275 peer-reviewed publications. She currently works on projects ranging from healthcare robots to developing methods to mitigate bias and trust in AI. She is a Fellow of IEEE, AAAI, AAAS, and the National Academy of Inventors (NAI). She was also recently elected to the American Academy of Arts and Sciences. To date, Dr. Howard’s unique accomplishments have been highlighted through a number of other public recognitions, including being recognized as one of the 23 most powerful women engineers in the world by Business Insider and one of the Top 50 U.S. Women in Tech by Forbes. In 2013, she also founded Zyrobotics, which developed STEM educational products to engage children of diverse abilities. Prior to Georgia Tech, Dr. Howard was at NASA's Jet Propulsion Laboratory where she held the title of Senior Robotics Researcher and Deputy Manager in the Office of the Chief Scientist. She also served on the National Academies’ Committee on Responsible Computing Research: Ethics and Governance of Computing Research and its Applications and currently serves on the National Artificial Intelligence Advisory Committee (NAIAC) and on the standing committee of the One Hundred Year Study on Artificial Intelligence (AI100).

The Secure Smart Grid: Cross-Layer, Intelligent, Cyber Physical Security for Smart Grid Systems

Tuesday, October 11, 2022 at 3:30PM EST

Dr. Janise McNair
Associate Professor
University of Florida

Biology

Dr. Janise McNair is an associate professor of electrical & computer engineering (ECE) at the University of Florida. She earned her B.S. and M.S. degrees from the University of Texas at Austin and her Ph.D. degree in ECE from the Georgia Institute of Technology, with a research focus on medium access control and mobility management in next generation wireless networks. She was a pioneer in early next generation mobility management for cellular systems and is currently a researcher in the development of multi-discipline, cross-layer networked systems. In her current research, she applies software-defined network (SDN) and machine learning principles to create adaptive telecommunications infrastructures for physical systems, including agriculture, remote construction sites, and smart grid. In addition, she creates network security protocols to protect these systems from attack. Dr. McNair is a member of the Intelligence, Science and Technology Experts Group of the National Academies of Sciences Engineering and Medicine, was a participant in the 2008 DARPA Computer Science Study Group, and currently serves on the IEEE Computer Society Integrity Committee, the editorial board of Nature Communications Engineering Journal and the Springer Wireless Networks Journal.

Abstract

Smart Grid research and development has drawn much attention from academia, industry, and government due to the great impact it will have on society, economics, and the environment. Securing the smart grid remains a key and complex problem consisting of physical process control and site monitoring to provide real-time protection from faults and attacks that disrupt the power system's operation. In this talk, Dr. McNair will discuss the security challenges of smart grid systems, and the development of a cross-layer suite of protocols for physical sensing, network monitoring, and software defined network management, fueled by machine learning algorithms and analysis. Adding a machine learning layer to the physical and network layers of a smart power grid enables use of knowledge of verified data to learn the normal state of a properly functioning grid and to detect anomalies introduced into the system. The Ensemble CorrDet protocol suite uses these adaptive statistics in the detection of bad data in power systems to account for the continually changing state of a power system. This provides an ability to understand normal and bad data behavior at both spatial levels (local detectors at each bus) and temporal levels (time varying adaptive threshold). Furthermore, it integrates the detection of faulty smart grid measurement data and inconsistent network packet behavior for more reliable and accurate anomaly detection and attack interpretation. This work is the first to consider multiple coordinated smart grid security attacks taking place at a single time. Lastly, the challenges of cross-layer system integration, simulation and modeling are discussed.

Explainability in Experimentation

Friday, September 30, 2022 at 11:00AM EST

Biology

Violet Syrotiuk is an associate professor in the School of Computing and Augmented Intelligence at Arizona State University. She studies complex engineered networks. Her interests include medium access control (MAC) protocols for wireless networks, dynamic adaptation of significant radio and/or protocol parameters to optimize performance for evolving network conditions, and petabyte-scale file transfer in programmable networks. Her focus is on implementations using mid-scale experimental infrastructure, such as the FABRIC testbed. Dr. Syrotiuk serves on the editorial boards of the Elsevier journals Computer Networks and Computer Communications, and on the technical program committees of major conferences sponsored by the IEEE and ACM.

Abstract

Understanding networked computing systems, such as the new FABRIC testbed, is crucial to harnessing their potential. On one hand, machine learning (ML) has been remarkably successful at building accurate predictive models of such complex systems, but explainability remains a challenge. On the other hand, traditional design of experiments (DoE) emphasizes systematic techniques for model building in terms of parameters and interactions among them used in experimentation, but it encounters difficulty in scalability. Adapting to dynamics in the system is a challenge for both approaches. Our interest is in the science of design and analysis of screening experiments to characterize large-scale engineered systems. The goal is to understand which and how parameters are interacting to inform the design of protocols for system operation and control. Whether by ML, DoE, or some middle course, in order to characterize a system and validate understanding, design of the data collection can be fundamental.

Panel on “Women in AI Networking – Industry Career”

Friday, June 3, 2022 at 3:00PM EDT

The panel will discuss the challenges and opportunities that females face during their careers in industry. The panelists are

• Dr. Sujata Banerjee, Vice president of VMware Research Group
• Dr. Renata Teixeira, Visiting Scholar at the Streaming Algorithm Team at Netflix
• Dr. Victor Bahl, Technical Fellow and Chief Technology Officer at Microsoft, Azure.

Panel on “Women in AI Networking – Academic Career”

Tuesday, March 8, 2022 at 1:00PM EST

The panel features a diverse group of outstanding female academics who discuss the challenges and opportunities that female professors face during their academic careers.

The panelists are

• Prof. Elisa Bertino, Purdue University
• Prof. Gauri Joshi, Carnegie Mellon University
• Prof. Mingyan Liu, University of Michigan
• Prof. Chunyi Peng, Purdue University
• Prof. Aylin Yener, Ohio State University