What? Formal and Efficient Analysis and Design of Complex Systems 

When? Tuesday, February 06, 2024 @ 11 a.m.

Where? CODA 9th floor Atrium 0920 or virtually on Zoom

Will there be free lunch? YES!

Additional information

Abstract: The task of system design is shared by all engineering disciplines, each coming  

with its own techniques. In spite of their differences in tools, there is large  

intersection in their conceptual approach to design. In this talk, we exploit  

this commonality to take an abstract view of systems and their composition. We  

understand systems and subsystems in terms of their assume-guarantee  

specifications, or contracts.  

Assume-guarantee contracts are formal specifications that state (i) the  

assumptions that a design element makes on its environment and (ii) the  

guarantees it delivers when the environment behaves according to the contract's  

assumptions. Contracts come with a rich algebra that allows us to carry out  

several design-relevant tasks: obtaining system-level specifications from  

component specifications, finding specifications of components that need to be  

added to a design in order to meet an objective, etc. We will introduce the  

algebra of contracts and discuss how the various algebraic operations relate to  

system-design tasks. We will discuss hypercontracts, an extension of  

assume-guarantee reasoning to support the formal analysis of key security and  

robustness properties. We will also discuss the application of this methodology  

and Pacti, a software package that supports system design using contracts, in  

applications ranging from space-mission design to synthetic biology.  

Bio: Inigo Incer is a postdoctoral researcher at Caltech and UC Berkeley. He obtained  

his PhD from UC Berkeley in 2022 under the guidance of Alberto  

Sangiovanni-Vincentelli. He is interested in all aspects of cyber-physical  

systems, emphasizing formal methods and AI that support their compositional  

design and analysis. Before pursuing a PhD, Inigo was an IC designer in Austin.  

His work has been supported by the NSF/ASEE eFellows program and the UC Berkeley  

Chancellor's Fellowship.  

Zoom Link: https://gatech.zoom.us/webinar/register/WN_I_NIDrazS-OFf0VqMOR4Vg

Date: Tuesday, February 6, 2024 at 11:00 am

Location: Coda Building, Second Floor, Room 230 (Google Maps link)

Link: The recording of this in-person seminar will be uploaded to CSE's MediaSpace

Title: Intelligent Software in the Era of Deep Learning

Abstract: With the end of Moore's Law and the rise of compute- and data-intensive deep-learning (DL) applications, the focus on arduous new processor design has shifted towards a more effective and agile approach -- Intelligent Software to maximize the performance gains of DL hardware like GPUs.

In this talk, I will first highlight the importance of software innovation to bridge the gap between the increasingly diverse DL applications and the existing powerful DL hardware platforms. The second part of my talk will recap my research work on DL system software innovation, focusing on bridging the 1) Precision Mismatch between DL applications and high-performance GPU units like Tensor Cores (PPoPP '21 and SC '21), and 2) Computing Pattern Mismatch between the sparse and irregular DL applications such as Graph Neural Networks and the dense and regular tailored GPU computing paradigm (OSDI '21 and OSDI '23). Finally, I will conclude this talk with my vision and future work for building efficient, scalable, and secure DL systems.

Bio: Yuke Wang is a final-year Doctor of Philosophy (Ph.D.) candidate in the Department of computer science at the University of California, Santa Barbara (UCSB). He got his Bachelor of Engineering (B.E.) in software engineering from the University of Electronic Science and Technology of China (UESTC) in 2018. At UCSB, Yuke is working with Prof. Yufei Ding (Now at UC at San Diego, CSE). Yuke's research interests include Systems & Compiler for Deep Learning and GPU-based High-performance Computing. His projects cover graph neural network (GNN) optimization and its acceleration on GPUs. Yuke’s research has resulted in 20+ publications (with 10 first-authored papers) in top-tier conferences, including OSDI, ASPLOS, ISCA, USENIX ATC, PPoPP, and SC. Yuke’s research outcome has been adopted for further research in industries (e.g., NVIDIA, OctoML, and Alibaba) and academia (e.g., University of Washington and Pacific Northwest National Laboratory). Yuke is also the recipient of the NVIDIA Graduate Fellowship 2022 (Top-10 out of global applicants) and has industry experience at Microsoft Research, NVIDIA Research, and Alibaba. The ultimate goal of Yuke’s research is to facilitate efficient, scalable, and secure deep learning in the future.

Date: Tuesday, February 20, 2024 at 11:00 am

Location: Coda Building, Second Floor, Room 230 (Google Maps link)

Link: The recording of this in-person seminar will be uploaded to CSE's MediaSpace

Title: Accessible Foundation Models: Systems, Algorithms, and Science

Abstract: The ever-increasing scale of foundation models, such as ChatGPT and AlphaFold, has revolutionized AI and science more generally. However, increasing scale also steadily raises computational barriers, blocking almost everyone from studying, adapting, or otherwise using these models for anything beyond static API queries. In this talk, I will present research that significantly lowers these barriers for a wide range of use cases, including inference algorithms that are used to make predictions after training, finetuning approaches that adapt a trained model to new data, and finally, full training of foundation models from scratch.  For inference, I will describe our LLM.int8() algorithm, which showed how to enable high-precision 8-bit matrix multiplication that is both fast and memory efficient. LLM.int8() is based on the discovery and characterization of sparse outlier sub-networks that only emerge at large model scales but are crucial for effective Int8 quantization. For finetuning, I will introduce the QLoRA algorithm, which pushes such quantization much further to unlock finetuning of very large models on a single GPU by only updating a small set of the parameters while keeping most of the network in a new information-theoretically optimal 4-bit representation. For full training, I will present SWARM parallelism, which allows collaborative training of foundation models across continents on standard internet infrastructure while still being 80% as effective as the prohibitively expensive supercomputers that are currently used. Finally, I will close by outlining my plans to make foundation models 100x more accessible, which will be needed to maintain truly open AI-based scientific innovation as models continue to scale.

Bio: Tim Dettmers’s research focuses on making foundation models, such as ChatGPT, accessible to researchers and practitioners by reducing their resource requirements. This involves developing novel compression and networking algorithms and building systems that allow for memory-efficient, fast, and cheap deep learning. These methods enable many more people to use, adapt, or train foundation models without affecting the quality of AI predictions or generations. He is a PhD candidate at the University of Washington and has won oral, spotlight, and best paper awards at conferences such as ICLR and NeurIPS. He created the bitsandbytes library for efficient deep learning, which is growing at 1.4 million installations per month and received Google Open Source and PyTorch Foundation awards.

Date: Thursday, February 8, 2024 at 11:00 am

Location: Coda Building, Second Floor, Room 230 (Google Maps link)

Link: The recording of this in-person seminar will be uploaded to CSE's MediaSpace

Title: Digital Genome Reconstruction with De Novo Assembly

Abstract: De novo genome assembly, a methodology that reconstructs the genome without relying on a reference, has been a central problem and remains one of the most challenging tasks in bioinformatics for four decades. With the advancement of long-read sequencing technologies, assembly algorithms now have the capacity to routinely reconstruct high-quality diploid human genomes at an affordable cost. This significant progress facilitates comprehensive research and clinical studies on even the most complex genomes, as well as paves the way to the personal genome era. In this talk, I will present a suite of algorithms that I built for de novo genome assembly, including hifiasm (Cheng et al, Nature Methods, 2021), hifiasm (Hi-C) (Cheng et al, Nature Biotechnology, 2022) and hifiasm (UL) (Cheng et al, Nature Methods, accepted, 2024). These algorithms are designed to generate optimal genome assemblies by integrating various data types. Furthermore, this talk will explore potential improvement to the de novo genome assembly methodology and its application in emerging research areas.

Bio: Haoyu Cheng is a Postdoctoral Scholar working with Dr. Heng Li at Dana-Farber Cancer Institute and Harvard Medical School. Previously he obtained his Ph.D. degree in Computer Science from the University of Science and Technology of China. His research primarily focuses on developing computational methods for genomic data, with a particular emphasis on de novo genome assembly and its applications. His hifiasm algorithm has been widely used and has already become the dominant long-read genome assembler since its publication in 2021. In under three years, hifiasm has garnered over 1,600 citations, and has been funded by the K99/R00 Pathway award from the National Institutes of Health (NIH).

Bio: Todd Murphey is a Professor of Mechanical Engineering in the McCormick School of Engineering and of Physical Therapy and Human Movement Sciences in the Feinberg School of Medicine, both at Northwestern University.  He received his Ph.D. in Control and Dynamical Systems from the California Institute of Technology.  His laboratory is part of the Center for Robotics and Biosystems, and his research interests include robotics, control, human-machine interaction, and emergent behavior in dynamical systems. He received the National Science Foundation CAREER award, was a member of the 2014-2015 DARPA/IDA Defense Science Study Group, and is a current member of the United States Department of the Air Force Scientific Advisory Board.

Abstract: Platforms, devices, and algorithms are increasingly weaponized to control and harass the most vulnerable among us. Some of these harms occur at the individual and interpersonal level: for example, abusers in intimate partner violence (IPV) use smartphones and social media to surveil and stalk their victims. Others are more subtle, at the level of social structure: for example, in organizations, workplace technologies can inadvertently scaffold exploitative labor practices. This talk will discuss my research (1) investigating these harms via online measurement studies, (2) building interventions to directly assist survivors with their security and privacy; and (3) instrumenting these interventions, to enable scientific research into new types of harms as attackers and technologies evolve. I will close by sharing my vision for centering inclusion and equity in digital safety, security and privacy, towards brighter technological futures for us all.

Bio: Emily Tseng is a PhD candidate in Information Science at Cornell University. Her research explores the systems, interventions, and design principles we need to make digital technology safe and affirming for everyone. Emily’s work has been published at top-tier venues in human-computer interaction (ACM CHI, CSCW) and computer security and privacy (USENIX Security, IEEE Oakland). For 6 years, she has served as a researcher-practitioner with the Clinic to End Tech Abuse, where her work has enabled specialized security services for over 600 survivors of intimate partner violence (IPV). Emily is the recipient of a Microsoft Research PhD Fellowship, Rising Stars in EECS, Best Paper Awards at CHI, CSCW, and USENIX Security, and third place in the Internet Defense Prize. She has additionally completed internships at Google and with the Social Media Collective at Microsoft Research. She holds a B.A. from Princeton University. 

Date: Thursday, January 25, 2024 at 11:00 am

Location: Coda Building, Second Floor, Room 230 (Google Maps link)

Link: The recording of this in-person seminar will be uploaded to CSE's MediaSpace

Title: Understanding and Improving High-Stakes Decision Making: People, Algorithms, and Design

Abstract: High-stakes evaluation problems – to estimate the quality of items or people – arise in many real-world applications such as admissions, grading, and hiring. My research focuses on understanding and improving these evaluation problems along the dimensions of accuracy, fairness, efficiency, and reliability. In this talk, I describe one line of my research on correcting human bias under different behavioral models. The first type of bias arises in a sequential setting such as sports competitions and court decisions, where an evaluator rates candidates in an online, irrevocable fashion. I propose a natural model for the evaluator's rating process that captures the lack of calibration inherent to this task, and conduct crowdsourcing experiments to support various facets of the model. I propose an efficient bias correction algorithm and show that it is information-theoretically optimal in terms of common ranking error metrics. I also briefly describe a second type of bias arising from people’s experiences that are irrelevant to the evaluation objective. For example, in teaching evaluation, students who receive higher grades are more positive towards their instructors. In such scenarios, I propose an adaptive algorithm that corrects student ratings under very mild non-parametric assumptions for the biases.

Bio: Jingyan Wang is a President's postdoctoral fellow in the H. Milton Stewart School of Industrial and Systems Engineering at the Georgia Institute of Technology. She received her Ph.D. from the School of Computer Science at Carnegie Mellon University, advised by Nihar Shah, and her B.S. in Electrical Engineering and Computer Sciences with a minor in Mathematics from the University of California, Berkeley. She uses tools from statistics and machine learning to understand and improve high-stakes decision-making systems such as those involving hiring and admissions. Her interdisciplinary research has been published in top venues in statistics, machine learning, artificial intelligence, human computation, and economics and computation. She is the recipient of the Best Student Paper Award at AAMAS 2019, and was selected as a Rising Star in EECS and in Data Science.

Date: Tuesday, January 23, 2024 at 11:00 am

Location: Coda Building, Second Floor, Room 230 (Google Maps link)

Link: The recording of this in-person seminar will be uploaded to CSE's MediaSpace

Title: Operationalizing Responsible Machine Learning: From Equality Towards Equity

Abstract: With the widespread proliferation of machine learning, there arises both the opportunity for societal benefit as well as the risk of harm. Approaching responsible machine learning is difficult because technical approaches may build on too many layers of abstraction, ending up prioritizing a mathematical definition of fairness that correlates poorly to real-world constructs of fairness. On the other hand, social approaches engaging with prescriptive theories may produce findings that are too abstract to translate well into practice. In my research, I bridge these approaches and use social implications to guide technical work. I will discuss three research directions that show how although the technically convenient thing to do is consider equality acontextually, through stronger engagement with societal context we can operationalize a more equitable formulation. First, I will introduce a dataset tool that we built to analyze complex, socially-grounded forms of visual bias. Then, I will give empirical evidence to support how we should incorporate societal context in bringing intersectionality into machine learning. Finally, I will talk about how to formulate the evaluation metric of bias amplification based on more realistic assumptions about the state of the world. Overall, I will cover how we can expand a narrow focus on equality in responsible machine learning to a broader understanding of equity that substantively engages with societal context.

Bio: Angelina Wang is a computer science Ph.D. student at Princeton University advised by Olga Russakovsky. Her research is in the area of machine learning fairness and algorithmic bias. She has been recognized by the NSF GRFP, EECS Rising Stars, and Siebel Scholarship. She has published in top machine learning (ICML, AAAI), computer vision (ICCV, IJCV), and responsible computing (FAccT, JRC) venues, including spotlight and oral presentations. Previously, she has interned with Microsoft Research and Arthur AI, and received a B.S. in Electrical Engineering and Computer Science from UC Berkeley.

Date: Thursday, February 1, 2024 at 11:00 am

Location: Coda Building, First Floor, Room 114 (Google Maps link)

Link: The recording of this in-person seminar will be uploaded to CSE's MediaSpace

Title: Solving Real-World Tasks with AI Agents

Abstract: For years, my dream has been to create autonomous AI agents capable of carrying out tedious procedural tasks (e.g., arranging conference travel), allowing me to focus on more creative and exciting tasks. Modern AI models, especially large language models (LLMs) like ChatGPT, have suddenly brought us much closer to achieving such AI agents. But, has my dream already come true? In this talk, I will answer this question by delving into my systematic evaluation of AI agents in realistic tasks. The evaluation uncovers many critical limitations of AI agents, such as accurate grounding, long-term planning, and tool use. It suggests that LLMs are crucial yet early steps towards AI autonomy. To address these challenges,  I will introduce my research of a more suitable “language” for AIs, which overcomes the inherent limitations of using natural language for task solving. Finally, I will discuss my work on teaching AI agents to learn new tools by reading the tool documentation rather than direct demonstrations.

Bio: Shuyan Zhou is a final-year Ph.D. student at the Language Technologies Institute at CMU, advised by Graham Neubig. Her research in NLP and AI focuses on creating AI agents for real-world tasks, such as using computers and generating code. Her work has been recognized at top natural language processing and machine learning conferences and journals such as ICLR, ICML, ACL, EMNLP, and TACL. You can find more about her at https://shuyanzhou.com

This talk discusses our recent efforts to bridge this gap. First, we show that by understanding the force responses from deformable terrains, we could allow robots to elicit desired ground reaction forces from challenging terrains like sand and mud and produce significantly improved locomotion performance. Second, we show that by leveraging the high force transparency of direct-drive actuators, robots could use their legs as proprioceptive sensors to determine substrate strength and mechanical properties. This proprioceptive sensing capability can enable robots to gather rich information from their environment during every step, and adapt their locomotion strategies accordingly. Finally, we discuss our latest progress in applying these locomotion and sensing strategies in earth and planetary exploration scenarios, and how the improved sensing and locomotion capabilities pave the way for new human-robot teaming workflows.

Bio: Feifei Qian is an Assistant Professor of Electrical and Computer Engineering at the University of Southern California. Qian received her PhD in Electrical Engineering and M.S. in Physics from Georgia Institute of Technology, in 2015 and 2011, respectively. Prior to her appointment at USC, she worked in the GRASP lab at University of Pennsylvania as a postdoctoral fellow. Qian’s expertise is in analyzing and modeling the complex interactions between robots and environments and developing innovative control and sensing strategies to improve robot mobility on challenging terrains. Qian’s research has been recognized with NSF CAREER award, best student paper award from the Robotics Science and Systems conference and has been featured in several media press including BBC News, Phys.org, and R&D Magazine.

The workshop is jointly sponsored by Georgia Tech's Institute for Data Engineering and Science and School of Computational Science and Engineering. 

Website: https://rpestourie.github.io/scientific-ai-gt2024/

Brief Description This one-day workshop will be held at Georgia Tech on January 16th, 2024 and convene experts in computational mathematics, statistics, data science, and application domain scientists across science and engineering to address the foundational topics of scientific AI for the prediction and optimization of complex systems. The agenda emphasizes transcending conventional forward simulations and predictions to realize the optimization of complex systems at a large scale by harnessing the power of scientific AI.

Schedule

Breakfast 8:15 AM - 8:45 AM: Srinivas Aluru & Haesun Park

Opening remarks 8:45 AM - 9:00 AM: Srinivas Aluru & Haesun Park

Session 1

• 9:00 AM - 9:35 AM: Prasanna Balaprakash, ORNL (30 mins + 5 mins Q&A)
  Scalable, Automated Deep Neural Network Development for Complex Systems
• 9:35 AM - 9:50 AM: Raphaël Pestourie, GT CSE (10 mins + 5 mins Q&A)
  Scientific Machine Learning Enables Design via Surrogate Models and End-to-End Training
• 9:50 AM - 10:05 AM: Haomin Zhou, GT Math (10 mins + 5 mins Q&A)
  Parameterized Wasserstein Geometric Flow

Morning Break 10:05 AM - 10:25 AM (20 mins)

Session 2

• 10:25 AM - 11:00 AM: Bart van Bloemen Waander, Sandia (30 mins + 5 mins Q&A)
  Hyper-differential Sensitivity Analysis with Respect to Model Discrepancy
• 11:00 AM - 11:15 AM: Amir Poursaee, Clemson (20 mins + 5 mins Q&A)
  Corrosion and AI
• 11:25 AM - 11:40 PM: Lars Ruthotto, Emory Math (10 mins + 5 mins Q&A)
  Generative Modeling for Simulation-Based Inference

Panel Discussion

• 11:40 AM - 12:30 PM: Panel Discussion (50 mins)
  Panelists: Prasanna Balaprakash, Bart van Bloemen Waander, Amir Poursaee, Lars Ruthotto
  Moderator: Raphael Pestourie

Lunch Break and Networking 12:30 PM - 1:30 PM (60 mins)

Session 3

• 1:30 PM - 2:05 PM: Panos Stinis, PNNL (30 mins + 5 mins Q&A)
  Multifidelity Scientific Machine Learning
• 2:05 PM - 2:20 PM: Bo Zhu, GT IC (10 mins + 5 mins Q&A)
  Computational Fluid Dynamics on Neural Flow Maps
• 2:20 PM - 2:35 PM: Bolei Deng, GT AE (10 mins + 5 mins Q&A)
  Inverse Design and Optimization of Mechanical Metamaterials
• 2:35 PM - 2:50 PM: Yan Wang, GT MechE (10 mins + 5 mins Q&A)
  Physics-Informed Machine Learning: (Over-)Promising?

Afternoon Break 2:50 PM -3:10 PM (20 mins)

Session 4

• 3:10 PM - 3:45 PM: Guannan Zhang, ORNL (30 mins + 5 mins Q&A)
  Generative Machine Learning Models for Uncertainty Quantification
• 3:45 PM - 4:10 PM: Aditi Krishnapriyan, UC Berkeley (20 mins + 5 mins Q&A)
  Exploring the Role of Physics-Constrained Machine Learning with Soft and Hard Constraints
• 4:10 PM - 4:25 PM: Peng Chen, GT CSE (10 mins + 5 mins Q&A)
  Derivative-Informed Neural Operators for PDE-Constrained Optimization under Uncertainty

Panel Discussion and closing 

• 4:25 PM - 5:30 PM: Panel Discussion (65 mins)
  Panelists: Panos Stinis, Bo Zhu,Guannan Zhang,Aditi Krishnapriyan
  Moderator: Peng Chen

Workshop Details

Complex systems are prevalent in a wide range of scientific and engineering disciplines. These systems incorporate numerous variables, some of which are controllable, some are subject to design specifications, and others may be latent or determined by either identified or unidentified processes. Accurate prediction of the system response is intrinsically difficult, yet often secondary to more pressing engineering concerns regarding optimal design or control of the system. For instance, in the realm of electromagnetic (EM) devices, advancements in fabrication technologies have afforded extensive design and control freedoms, such as metamaterials and 4D materials. These developments have rendered simulation and optimization as the limiting factors for inverse design or control applications. In the context of corrosion science, the corrosion initiation of metals is governed by dynamic and complex processes that take place at multi-scale levels. As such, investigation of its kinetic dissolution with just macroscopic, or bulk, electrochemical techniques or sometimes with high resolution microscopy poses considerable challenges. Scientific AI offers a viable avenue for the rapid and reliable prediction and discovery during the initiation and propagation of metallic corrosion. The intersection of scientific AI and complex systems engenders a host of unresolved questions, including efficient simulation strategies, model parameterization, the integration of domain-specific knowledge or system architecture, and the role of AI in scientific computing for predictive and optimization purposes.

The methodologies of Scientific AI utilize the robust nonlinear fitting capabilities of deep learning to enhance the predictive efficiency and accuracy of conventional scientific computing and statistical methods for the simulation of complex systems. This workshop aims to explore how scientific AI employs domain-specific knowledge and scientific computational techniques to improve the accuracy and efficiency of the optimization of complex systems. It is essential to note that a scientific AI model proficient in making accurate predictions is not invariably well-suited for optimization tasks. Therefore, this workshop will concentrate on approaches that facilitate effective optimization performance of scientific AI, whether through gradient-based optimization techniques or alternative strategies, contributing to groundbreaking scientific and engineering discoveries through optimization.

Organizers: Raphaël Pestourie, Peng Chen, Ebeneser Fanijo, Yan Wang

Due to venue size, this workshop is invitation only. Please contact one of the organizers if you are interested and not a current invitee.

This event is being co-sponsored by Georgia Tech's MS in Human-Computer Interaction (MS-HCI) and the Institute for People and Technology (IPaT). MS-HCI students, Ph.D. students, and faculty will engage in people-centric innovation during the research showcase and networking event.

The event will be held Tuesday, Feb. 6, at the Technology Square Research Building.

Karen Holtzblatt will serve as keynote speaker with her talk, What is the Actual UX Job? Strategy, Process Leadership, Listening and Communicating, and Organizational Management. 

AGENDA

12:30 - 1:00 PM   Registration

1:00 - 2:00 PM     Keynote: Karen Holtzblatt: What is the Actual UX Job?
Strategy, Process Leadership, Listening and Communicating, and Organizational Management

2:00 - 3:00 PM    One-Minute Madness – MS-HCI students present their research and interests to industry representatives

3:00 - 5:00 PM    Research posters and demonstrations

4:30 - 6:00 PM    Reception + Networking – Industry guests, students, faculty,  and researchers

Registration:  https://www.eventbrite.com/e/interactivitygt-tickets-754656988367?aff=oddtdtcreator

Title: Deep Learning on Graphs: A Data-Centric Exploration

Abstract: Many learning tasks in Artificial Intelligence require dealing with graph data, ranging from biology and chemistry to finance and education. Graph neural networks (GNNs), as deep learning models, have shown exceptional capabilities in learning from graph data. Despite their successes, GNNs often grapple with challenges stemming from data size and quality. This talk emphasizes a data-centric approach to enhance GNN performance. First, I will demonstrate methods to significantly reduce graph dataset sizes while retaining essential information for model training. Next, I will introduce a model-agnostic framework that enhances the quality of imperfect input graphs, thereby boosting prediction performance. These data-centric strategies not only enhance data efficiency and quality but also complement existing models. Finally, I will introduce recent advances in graph data valuation and graph generation. Join us to explore innovative approaches for overcoming data-related challenges in graph data mining.

Bio: Wei Jin is an Assistant Professor of Computer Science at Emory University. He obtained his Ph.D. from Michigan State University in 2023. His research focuses on graph machine learning and data-centric AI, with notable accomplishments such as AAAI New Faculty Highlights, KAUST Rising Star in AI, Snap Research Fellowship, Most Influential Papers in KDD and WWW by Paper Digest, and top finishes in three NeurIPS competitions. He has organized tutorials and workshops at top conferences, and published in top-tier venues such as ICLR, KDD, ICML, and NeurIPS. He has served as (senior) program committee members at these conferences and received the WSDM Outstanding Program Committee Member award. 

Bio: Masayoshi Tomizuka received his B.S. and M.S. from Keio University in 1968 and 1970, respectively. He received his Ph. D. from MIT in 1974, after which he joined the ME Department at UC Berkeley. Here, he served as the Vice Chair of Instruction from Dec. 1989 to Dec. 1991, and as the Vice Chair of graduate studies from Jul. 1995 to Dec. 1996.

He is currently the Associate Dean of Academic Affairs for the College of Engineering at UC Berkeley. From 2009 to 2011, he was the Executive Associate Dean for the College of Engineering at UC Berkeley. He also served as Program Director of the Dynamic Systems and Control Program at the National Science Foundation from Sept. 2002 to Dec. 2004.

Prof. Tomizuka’s research interests include optimal and adaptive control, digital control, signal processing, motion control, mechatronics and their applications in robotics, manufacturing, data storage devices, vehicles, and human-machine systems.

Bio: I am an assistant professor in the Computer Science Department at Brown University. The aim of my research program is to construct robots that seamlessly use natural language to communicate with humans. In twenty years, every home will have a personal robot which can perform tasks such as clearing the dinner table, doing laundry, and preparing dinner. As these machines become more powerful and more autonomous, it is critical to develop methods for enabling people to tell them what to do. Robots that can communicate with people using language can respond appropriately to commands given by humans, ask questions when they are confused, and request help when they get stuck. We apply probabilistic methods, corpus-based training, and decision theory to develop interactive robotic systems that can understand and generate natural language. I completed my Ph.D. at the MIT Media Lab in 2010, where I developed models for the meanings of spatial prepositions and motion verbs. My postdoctoral work at MIT CSAIL focused on creating robots that understand natural language. I have published at SIGIR, HRI, RSS, AAAI, IROS, and ICMI, winning Best Student Paper at SIGIR and ICMI. I was named one of IEEE Spectrum’s AI’s 10 to Watch and won the Richard B. Salomon Faculty Research Award at Brown University.

Abstract

Hardly a day passes without a new technology ethics scandal in the news — from privacy violations on social media to biased algorithms to controversial data collection practices. In computing practice and research, good intentions sometimes still lead to negative consequences. So what can we do as technologists, researchers, and educators? This talk describes three lessons from my research that inform ethical practices in studying, building, and teaching about technology: (1) remembering the humans present in data, towards ethical research practices; (2) unpacking ethical debt (as a parallel to technical debt) in technology design and research as the precursor to the types of unintended consequences that underly many controversies; and (3) a broader perspective on computing education that puts thoughtful critique of technology in everyone’s hands. 

Bio

Casey Fiesler is an Associate Professor of Information Science (and Computer Science by courtesy) at University of Colorado Boulder. She researches and teaches in the areas of technology ethics, internet law and policy, and online communities. Her work on research ethics for data science, ethics education in computing, and broadening participation in computing has been supported by the National Science Foundation, and she is the recipient of an NSF CAREER Award. Also a public scholar, she is a frequent commentator and speaker on topics of technology ethics and policy, and her research has been covered everywhere from The New York Times to Teen Vogue (though she’s particularly proud of her TikToks). She is a three-time alum of Georgia Tech, and in addition to her PhD in Human-Centered Computing, she also holds a JD from Vanderbilt Law School.

Abstract

There is an inevitable gap between computational systems, which follow principles of precision, formalization, and abstraction, and human needs in the social world, which are contextual, fluid, and nuanced. Making AI technologies more responsible and human-centered requires understanding and bridging this socio-technical gap. Focusing on my research on explainable AI (XAI), and more recently large language models (LLMs), I will discuss four approaches my collaborators and I take to bridge the socio-technical gap for emerging and consequential AI technologies: conducting critical investigations into dominant AI algorithmic paradigms; researching and supporting AI practitioners who address the socio-technical gap on the ground; developing socially-guided computation and design frameworks; and shaping technical development by embedding socio-technical perspectives in the evaluation practices. I will conclude the talk with future directions toward my long-term research agenda of informing and developing AI technologies that are beneficial and adaptable for different stakeholder groups and usage contexts.  

Q. Vera Liao is a Principal Researcher at Microsoft Research Montréal, where she is part of the FATE (Fairness, Accountability, Transparency, and Ethics in AI) group. Her current research interests are in human-AI interaction, explainable AI, and responsible AI, with an overarching goal of bridging emerging AI technologies and human-centered perspectives. Prior to joining MSR, she worked at IBM Research, and studied at the University of Illinois at Urbana-Champaign and Tsinghua University. She has authored more than 60 peer-reviewed research articles and received many paper awards at HCI and AI venues. She currently serves as the co-editor-in-chief for the Springer HCI Book Series, in the Editors team for ACM CSCW conference, and on the Editorial Board of ACM TiiS. She has also served as an Area Chair and Senior PC member for CHI, FAccT, and IUI conferences. 

Title: Trustworthy Optimization Learning

Abstract: The fusion of deep learning and optimization has the potential to deliver outcomes for engineering applications that the two technologies cannot achieve independently. This talk illustrates this potential with the concept of optimization proxy, a differentiable program that can produce, in milliseconds, feasible and near-optimal solutions to classes of optimization problems. The talk reviews some of the foundations underlying optimization proxies, including end-to-end learning, compact optimization learning, dual learning, and self-supervised learning. The benefits of optimization proxies are demonstrated on applications in power systems and supply chains.

Bio: Pascal Van Hentenryck is the director of the NSF AI Institute for Advances in Optimization (AI4OPT) and the A. Russell Chandler III Chair and Professor at Georgia Tech. He was a professor at Computer Science at Brown University for over 20 years and led the optimization research group at the National ICT Research, Australia. Van Hentenryck designed and implemented several innovative optimization systems that have been in commercial use for over 20 years, including the pioneering CHIP and OPL systems. His current research focuses on AI Engineering, for applications in energy systems, supply chains, manufacturing, and mobility.

Title: TBD

Abstract: TBD

Bio: TBD

Title: TBD

Abstract: TBD

Bio: TBD

Title: TBD

Abstract: TBD

Bio: TBD

Bio: Kevin Lynch is a professor of mechanical engineering and director of the Center for Robotics and Biosystems at Northwestern University. His research is on robotic manipulation, locomotion, human-robot systems, and robot swarms. He is former Editor-in-Chief of the IEEE Transactions on Robotics and the IEEE Conference on Robotics and Automation Conference Editorial Board, a coauthor of three textbooks on robotics and mechatronics, and the instructor of six Coursera online courses and the associated YouTube videos forming the Modern Robotics specialization. He received the B.S.E. degree in electrical engineering from Princeton University and the Ph.D. degree in robotics from Carnegie Mellon University.

Bio: TBA

Bio: Dora Sadigh is Assistant Professor in the Computer Science Department and Electrical Engineering Department at Stanford University and Director of Intelligent and interactive Autonomous Systems Group. Her research interests lie at the intersection of robotics, machine learning, and control theory. Specifically, her group is interested in developing efficient algorithms for safe, reliable, and adaptive human-robot and generally multi-agent interactions.

This event celebrates Georgia Tech book publications, showcasing the range and depth of scholarship on our campus. The 2024 event honors authors and editors who have published books between January 2022 and December 2023. 

Title: Towards Simulations of Complex Physics in High-speed Propulsion Devices Using High-Order Methods

Abstract: High-speed combustion devices involve intricate and interrelated physical phenomena that present significant challenges for system design analyses. These systems center around compressible turbulent airflows, often at supersonic velocities, interacting with fuel at time scales comparable to the chemical reactions driving combustion. In practical systems, various factors, including heat transfer to the combustor walls and the fuel source's thermochemistry, can markedly influence this process, leading to operational envelopes highly sensitive to system design. These complexities are challenging to replicate and study experimentally, underscoring the importance of numerical simulations in comprehending high-speed combustion environments. However, employing these tools for design purposes remains problematic. First-principles simulations can be cost-prohibitive, while designing and calibrating reduced-order models can be intricate. Moreover, the required level of fidelity to model such intricate flows can vary from case to case, often not known until the investigation begins. Consequently, there is an ongoing imperative to gain a profound understanding of the fundamental physics in these complex combustion settings.

This presentation will spotlight the continuous efforts of the Naval Research Laboratory to advance numerical simulation methods for a more comprehensive grasp of the intricate physics within high-speed combustion devices. Particular emphasis will be placed on the utilization of the discontinuous Galerkin methodology in conjunction with realistic physical models for combustion processes. The discussion will also offer a brief historical perspective to provide context regarding the current state of the art, identify existing scientific challenges, and outline future research opportunities.

Bio: Dr. Ryan Johnson's joined NRL in 2015 as Karles' Fellow, dedicating himself to advancing the JENRE® Multiphysics Framework by embedding intricate combustion models. His responsibilities included overseeing the integration of gas-phase reactions with finite rate kinetics for modeling efforts involving solid, liquid, and gaseous fuels. Dr. Johnson's current focus centers on the practical application and validation of the JENRE® Multiphysics Framework in Navy-relevant combustors, including diverse applications such as solid fueled ramjets, scramjets, and rotating detonation engines. Before his tenure at NRL, Dr. Johnson earned his PhD at the University of Virginia under the guidance of Professor Harsha Chelliah. His doctoral research explored the coupling of computational fluid dynamics (CFD) with theoretical chemically reacting flow, covering a broad spectrum from carbon surface reactions to hypersonics. His academic journey was supported by several fellowships, including the National Defense Science and Engineering Graduate (NDSEG) Fellowship, the Farrah Fellowship, and the Paul Voight Teaching Fellowship.

Dr. Johnson is interested and devoted to introducing high-fidelity models and innovative numerical techniques as alternatives to conventional methods for simulating chemically reacting flows. His dedication to advancing the field has been acknowledged with two prestigious NRL ARPAD research publication awards.

Title: Toward Multi-Modal, Multi-Source Foundation Models to Advance Large-Scale Neural Data Analysis

Abstract: Despite the rapid growth of datasets in neuroscience, models of brain function often struggle to harness this expansive scale due to variability across experiments and shifts in our measurements of the brain's state. In this talk, I will outline our efforts to consolidate diverse datasets from various brain regions and experimental conditions into a cohesive "neurofoundation" model. By leveraging extensive pretraining on this neural data, we aim to enable robust generalization across different modalities and species. Further, I will delve into how this foundational model promises to improve data efficiency, expand the capabilities of brain-machine interfaces and neural decoders, and provide advanced, user-friendly tools to the wider neuroscience community, setting the trajectory for a more integrated approach to neural data analysis.

Bio: Eva Dyer (she/they) is an Associate Professor in the Coulter Department of Biomedical Engineering at the Georgia Institute of Technology. Dr. Dyer’s research combines artificial intelligence (AI) and neuroscience to understand brain function (AI for Neuro) and to build abstractions of biological organization and function that can be used to create more flexible AI systems (Neuro for AI). Eva completed all of her degrees in Electrical & Computer Engineering, obtaining her Ph.D. and M.S. from Rice University and a B.S. from the University of Miami. Eva is the recipient of a number of awards, including a Sloan Fellowship in Neuroscience, NSF CAREER Award, Next Generation Leader Award from the Allen Institute, a McKnight Foundation Technological Innovations in Neuroscience Award, and a CIFAR Azrieli Global Scholar Award.

Presenter’s Name: Heidi Hysell

Room: Virtual (Link will be sent the day before the event)

Time: 12 p.m. – 1 p.m. ET

Date: November 16

Abstract: Let's cut to the chase—entrepreneurship and innovation are more than just buzzwords; they are mechanisms for positive change. In this talk, I'll share a roadmap for those thinking of founding or are in the middle of running a startup. We'll go through the critical phases and touch on topics such as idea validation, funding, scaling, and long-term impact. We'll go beyond theory and hear stories from practice. We'll talk about how we can build better futures; laying the groundwork today for a more innovative, responsible, and inclusive tomorrow. 

Bio: 

Heidi Hysell is the founder of Future Founders and works with founders and CxOs to build better businesses and better futures. Her insights, frameworks, and support enable organizations to envision the future and evolve businesses. With a track record that includes advising portfolio companies for Samsung Next, launching Alpha Drive, a pioneering startup in benchmarking embodied AI, and consulting with Fortune 500 firms with Board of Innovation, she's seen firsthand what it takes to succeed in this game and have an impact. 

RSVP Form: https://gatech.co1.qualtrics.com/jfe/form/SV_bIbcAoIyZ85w4vQ

Title: A Systems Based Data Science Approach to Studying Acute and Critical Illness

Abstract: This talk will cover the challenges and opportunities in utilizing multi-modal data, often consisting of irregular time series, observational and experimental data, to develop predictive models of complex and dynamic clinical course among critically ill patients in the intensive care unit (ICU). The ICU is a data intensive environment, which contains a large number of devices and interfaces that continuously capture data streams. However much of this data is curated based on the clinical suspicion of the care provider, thus patients presenting with the same condition may have multiple permutations of data collection. In this talk, I highlight the challenges and opportunities in modeling complex problems related to critical illness and the approaches we utilize to better understand how data can inform clinical decision making at the bedside.

Bio: Rishi Kamaleswaran is an Associate Professor at Emory University, Department of Biomedical Informatics, with secondary appointments in Pediatrics and Emergency Medicine. He earned his Ph.D. in Computer Science from the University of Ontario Institute of Technology in Canada. He is the Director of Translational Informatics within the School of Medicine and the Georgia CTSA. He is the Co-Director of the NIH P30 Georgia Cystic Fibrosis Research and Translation Core Center where he oversees a number of clinical informatics research. He also oversees the Emory Real-Time Data Science and Decision Support (RADS2) center that advances AI/ML applications at the bedside. His research is funded by a number of federal and regional agencies including the National Institutes of Health, Veteran's Affairs, Department of Defense, Michael J. Fox Foundation, and the CF foundation.

Title: Examining Shared Structure Across Multivariate Datasets

Abstract: A number of recent studies have sought to understand the behavior of artificial and biological neural networks by comparing the structure of representations across layers, networks and brain areas. Increasingly prevalent, too, are comparisons across modalities of data, such as neural network activations and training data or behavioral data and neurophysiological recordings. Many of these studies use dimensionality as a proxy for computational complexity, measured by approaches such as canonical correlation analysis (CCA), but a standard, model-free measure is lacking. We present a candidate measure for shared dimensionality that we call the effective number of shared dimensions (ENSD).  We illustrate how the ENSD and its constituent statistics can be applied in a variety of analyses, including on across-layer similarities of neural network activations and communication subspaces between brain regions. By comparing it to existing metrics we find that the ENSD can be re-written as a scaled version of centered kernel alignment (CKA) that additionally describes the dimensionality of the aligned subspace and, unlike CCA, is robust to cases where data is sparse or low rank. We also include a novel analysis on the organization of the fly olfactory system, where the ENSD provides insight into the interaction between innate and learned olfactory representations. Altogether, we present an interpretable and computationally efficient model-free measure of shared dimensionality that can be used to probe shared structure in a wide variety of data types.

Bio: Dr. Aoi is an Assistant Professor of Data Science and Neurobiology at the University of California, San Diego. His work is aimed at the development of statistical and machine learning techniques for the analysis of neural population activity with the aim of determining the computational principles of biological neural networks. 

Topic: Inverse problems in seismic imaging

Lunch will be provided at the event. Please fill out the RSVP form in the link below to let us know your lunch preferences.

RSVP Link: https://forms.office.com/r/aSamAxTShz

Bio: Michael Posa is an Assistant Professor in Mechanical Engineering and Applied Mechanics at the University of Pennsylvania. He leads the Dynamic Autonomy and Intelligent Robotics (DAIR) lab, a group within the Penn GRASP laboratory.  His group focuses on developing computationally tractable algorithms to enable robots to operate both dynamically and safely as they interact with their environments. Michael received his Ph.D. in Electrical Engineering and Computer Science from MIT in 2017, where, among his other research, he spent time on the MIT DARPA Robotics Challenge team. He received his B.S. in Mechanical Engineering from Stanford University in 2007. Before his doctoral studies, he worked as an engineer at Vecna Robotics. He received the NSF CAREER Award in 2023, the RSS Early Career Spotlight in 2023, a Google Faculty Research Award, and a Young Faculty Researcher Award from the Toyota Research Institute. His work has also received awards recognition at TRO, ICRA, Humanoids, and HSCC.

Bio:  Fei Xia is a Senior Research Scientist at Google DeepMind where he works on the Robotics team. He received his PhD degree from the Department of Electrical Engineering, Stanford University. He was co-advised by Silvio Savarese in SVL and Leonidas Guibas. His mission is to build intelligent embodied agents that can interact with complex and unstructured real-world environments. His research has been awarded a CoRL Special Innovation award and an ICRA Outstanding Robot Learning Paper Award, and featured on popular media outlets such as New York Times, Reuters, and WIRED. Most recently, He has been exploring using foundation models for robot decision making.

Title: The Power of Less: Harnessing Sparsity for Performance Optimization

Abstract: Sparse matrix computations are fundamental to scientific computing and data analytics applications, such as computer graphics and machine learning. Sparsity leads to irregular memory accesses that pose challenges for code optimization. While  specialized libraries can accelerate sparse computations, costly manual tuning is required for each application and architecture, reducing programmer productivity. Additionally, in machine learning, the unstructured sparsity patterns of deep learning models have rendered many of these libraries useless, prompting practitioners to use dense routine calls. Automation approaches such as compilers and runtime systems provide portability and ease of programming, but efficiently optimizing sparse codes remains a challenge due to access pattern irregularities.

In this talk, I will introduce our work on building compilers and automation frameworks to accelerate sparse numerical kernels. I will present a class of inspection strategies that automatically analyze information such as matrix sparsity patterns and the numerical methods' properties, to generate optimized sparse codes. Additionally, I will discuss algorithmic modifications that make machine learning and graphics applications more amenable to sparse code specialization and the use of sparse compilers.

Bio: Maryam Mehri Dehnavi is an Associate Professor at the Department of Computer Science at the University of Toronto, where she also serves as the Associate Chair of Research. She holds the prestigious Canada Research Chair in Parallel and Distributed Computing and is the recipient of  the Ontario Early Researcher Award. Additionally, she has served as the General Chair of PPoPP and is the Associate Editor for the Journal of Parallel and Distributed Computing. Maryam's works on building compilers and systems  that automatically transform computation patterns to patterns that are amenable  to optimization. Her software packages have been widely adopted by domain experts, particularly in computer graphics and machine learning. 

Title: Data Science for Scientific Data Compression and Transportation Applications

Abstract: In this talk, I will present my recent work on developing data science solutions for large scale applications in scientific computing and transportation.

The volume of data generated by scientific applications has continued to outpace the growth of computing power and storage capacities. Unlike video and image compression, scientists require downstream quantities of interest to be conserved along with error bounds on primary data. We have developed novel constrained auto encoders based algorithmic and software pipelines that satisfy both these requirements. For data generated using Tokomak simulation codes that produce petabytes of data using thousands of GPUS, we have shown that our approach can compress the data by two orders of magnitude while requiring less than one percent of computational resources.

Mitigating traffic congestion and improving safety are the cornerstones of transportation. We have developed an ML based edge-cloud based system to analyze and fuse data streams from LIDAR, cameras and ground sensor data for quantifying and impacting pedestrian and vehicle safety. Using the data collected, we are also developing digital twins that model behavior at traffic intersections. The underlying models are useful for detecting interruptions as well as optimizing signal timing plans. We will present experimental results on data collected from real-traffic intersections in the City of Gainesville and City of Orlando.

This work is supported by DOE, NSF, and FDOT.

Bio: Sanjay Ranka is a Distinguished Professor in the Department of Computer Information Science and Engineering at the University of Florida. From 1999-2002, as the Chief Technology Officer at Paramark (Sunnyvale, CA), he developed a real-time optimization service called PILOT for marketing campaigns. PILOT served more than 10+ million optimized decisions a day in 2002 with a 99.99% uptime. Paramark was recognized by VentureWire/Technologic Partners as a Top 100 Internet technology company in 2001 and 2002 and was acquired in 2002. Sanjay has also held positions as a tenured faculty member at Syracuse University, academic visitor at IBM and summer researcher at Hitachi America Limited.

He has coauthored one book, four monographs, 340+ journal and refereed conference articles. His recent coauthored work has received a best student paper runner-up award at IGARSS 2015, best paper award at BICOB 2014, best student paper award at ACM-BCB 2010, best paper runner-up award at KDD-2009, a nomination for the Robbins Prize for the best paper in the Journal of Physics in Medicine and Biology in 2008, and a best paper award at ICN 2007. His work has received 15,900+ citations with an h-index of 63 (based on Google Scholar).

He is a fellow of the IEEE, AAAS and AAIA (Asia-Pacific Artificial Intelligence Association) and a past member of IFIP Committee on System Modeling and Optimization. He received the 2020 Research Impact Award from IEEE Technical Committee on Cloud Computing and the 2022 Distinguished Alumnus Award from Indian Institute of Technology, Kanpur. He is a board member of American Society on Artificial Intelligence.

He is currently an associate editor for ACM Computing Surveys, IEEE/ACM Transactions on Computational Biology and Bioinformatics, Sustainable Computing: Systems and Informatics, Knowledge and Information Systems, and International Journal of Computing. Additionally, he is a book series editor for CRC Press for Bigdata. In the past, he has been an associate editor for IEEE Transactions on Parallel and Distributed Systems and IEEE Transactions on Computers, and an associate editor-in-chief of the Journal of Parallel and Distributed Computing. He was a general co-chair for ICDM in 2009, International Green Computing Conference in 2010, International Green Computing Conference in 2011, ACM Conference on Bioinformatics Computational Biology in 2012, and IEEE DataCOM in 2017. He was a program chair for 2013 International Parallel and Distributed Processing Symposium, 2015 High Performance Computing Conference, and a program co-chair for ICDM 2022.

He has consulted for several startups and Fortune 500 companies.

Date: October 24, 2023

Time: 11am – 1pm 

Location: 

Klaus Advanced Computing Building 

1116 East & West Seminar Room 

266 Ferst Drive, NW 

Atlanta GA 30332 

Bio: Scott Delp, Ph.D., is the James H. Clark Professor of Bioengineering, Mechanical Engineering, and Orthopaedic Surgery at Stanford University. He is the Founding Chairman of the Department of Bioengineering at Stanford and Director of the Wu Tsai Human Performance Alliance, which aims to transform human health through the science of peak performance. Scott is also the Director of the RESTORE Center, a NIH national center focused on measuring real world rehabilitation outcomes and Director of the Mobilize Center, a NIH National Center of Excellence focused on Big Data and Digitial Health. Scott’s laboratory develops technologies to advance movement science and human health. Software tools created in his lab, including OpenSim and Simtk.org, have become the basis of an international collaboration involving thousands of scientists who exchange simulations of human movement. He has published over 250 research articles and has recently released a book from MIT Press entitled Biomechanics of Movement: The Science of Sports, Robotics, and Rehabilitation. Professor Delp has co-founded six health technology companies and is a member of the U.S. National Academy of Engineering./p>

Bio: Dr. Oliver Brdiczka is currently serving as the Director of Applied Science at Adobe Inc., where he spearheads the integration of Generative AI technologies into Adobe’s suite of products, including Adobe Firefly. Prior to his tenure at Adobe, Dr. Brdiczka ran his own AI startup, and also made significant contributions to groundbreaking DARPA programs such as PAL and ADAMS during his time at Xerox PARC. Dr. Brdiczka’s research focus lies at the intersection of artificial intelligence and human-centered computing. Over the years, he has shared his findings and insights through more than 100 peer-reviewed research papers and has been granted over 40 patents.

Title: Performance-Guaranteed Learning and Control for Sustainable Energy Systems

Abstract: The past decade has witnessed success of learning and control methods such as reinforcement learning (RL) in a broad spectrum of applications, such as game play, robotics, and autonomous driving. As a result, the application of RL in sustainability applications has attracted surging attention recently. Despite the promise, one of the biggest challenges for the deployment of RL in energy systems is the lack of stability and performance guarantees. Since energy systems are critical infrastructure, failure to maintain stability can lead to catastrophic consequences.

In this talk, I will introduce our recent work on developing a stability constrained RL framework that combines policy learning in RL with Lyapunov stability in control to learn policy with formal stability guarantees. Using two illustrative sustainable energy system examples, I will discuss how can we systematically construct the stability constraints and incorporate them into RL via neural network structure design.

Bio: Yuanyuan Shi is an Assistant Professor at the Department of Electrical and Computer Engineering at the University of California San Diego. She received her Ph.D. in Electrical Engineering, masters in Electrical Engineering and Statistics, all from the University of Washington, in 2020. From 2020 to 2021, she was a Postdoctoral Scholar at the California Institute of Technology. Her research interests lie in the areas of machine learning, dynamical systems and control, with applications in sustainable energy systems. She is a recipient of multiple awards, including the Rising Star award in EECS by MIT in 2018, the Scientific Achievement Award from the University of Washington Clean Energy Institute in 2020, and the best paper finalist from ACM e-Energy 2022.

Bio: Meeko Oishi received the Ph.D. (2004) and M.S. (2000) in Mechanical Engineering from Stanford University (Ph.D. minor, Electrical Engineering), and a B.S.E. in Mechanical Engineering from Princeton University (1998). She is a Professor of Electrical and Computer Engineering at the University of New Mexico. Her research interests include human-in-the-loop control, stochastic optimal control, and autonomous systems. She previously held a faculty position at the University of British Columbia at Vancouver, and postdoctoral positions at Sandia National Laboratories and at the National Ecological Observatory Network. She was a Visiting Researcher at AFRL Space Vehicles Directorate, and a Science and Technology Policy Fellow at The National Academies. She is the recipient of the NSF CAREER Award, the NSF BRITE Fellowship, the Truman Postdoctoral Fellowship in National Security Science and Engineering, and a member of the 2022-2024 US Defense Science Study Group.

Title: Involution and Invariant-Domain Preserving Approximation of the Ideal Magneto-Hydrodynamics Equations

Abstract: I will discuss recent results regarding the approximation of Maxwell's equations written in first-order form. I will show in particular that the discontinuous Galerkin approximation is spectrally correct. This result is based on the observation that the discontinuous Galerkin approximation preserves a fundamental involution property which is known as Gauss's law in the literature. A similar result is conjectured to hold true for stabilized continuous finite elements. I will also present a numerical technique for solving the ideal MHD equations that is invariant-domain preserving and involution preserving. The key novelty is that no divergence cleaning is needed.  The method will be numerically illustrated using continuous finite elements. The first part of this project is a joint work with Alexandre Ern (ENPC, Paris).  The second part of this project is a joint work with Tuan Anh Dao (Uppsala University) and Ignacio Tomas (Texas Tech University).

Bio: Dr. Jean-Luc Guermond joined Texas A&M’s Department of Mathematics in 2004 and has been the Richard E. Ewing ExxonMobil Chair since 2012. Dr. Guermond’s research interests include numerical analysis and partial differential equations. He received his Ph.D. from the University of Paris VI before becoming a research associate in the Department of Ocean Engineering at MIT and the Computer Science Laboratory for Mechanics and Engineering Sciences in Orsay, France, amongst others. Dr. Guermond has served as International Chair for Inria and received grants from the Air Force Office of Scientific Research, Army Research Office, Department of Energy, and National Science Foundation. His is a Fellow of both the American Mathematical Society and the Society of Industrial and Applied Mathematics. He is a managing editor for various journals.

Title: Robust and Practical Solution of Laplacian Equations by Approximate Elimination

Abstract: In this talk, I will describe a new algorithm for solving Laplacian (and SDDM) linear equations. We introduce a variant of the approximate Cholesky factorization of Kyng and Sachdeva (FOCS 2016), and we experimentally demonstrate that this method performs well in practice across a broad range of problem classes. Our experiments suggest that our solver is much more robust than existing solvers for Laplacian linear equations, while retaining good performance across all instances.

Our factorization approach is simple: we eliminate matrix rows/columns one at a time and use sampling to update the remaining entries of the matrix, approximating complete Cholesky factorization. Unlike earlier approaches, our sampled entries always maintain a connected support graph on the neighbors of the eliminated variable.

This talk is based on joint work with Yuan Gao and Daniel Spielman.

Preprint: https://arxiv.org/abs/2303.00709

Bio: Rasmus Kyng is an assistant professor of theoretical computer science at ETH Zurich. Before joining the ETH CS department in 2019, he was a postdoctoral fellow at Harvard SEAS and the Simons Institute at UC Berkeley. In 2017, he received a PhD in computer science from Yale University, where he was advised by Daniel Spielman.

His work focuses on fast algorithms for solving graph problems, convex optimization, and structured linear equations. In addition, he develops hardness results for basic algorithmic problems and probability theoretic tools for algorithmic applications. His research has won the FOCS'22 best paper and FOCS'17 best student paper awards, and the ICBS 2023 Frontiers of Science Award for Best Paper in Theoretical Computer Science 2018-2022.

Support for travel expenses (up to $750 travel plus up to two nights hotel stay) will be provided for accepted participants. There are no citizenship requirements. Both prospective PhD and prospective MS students are welcome to apply; some priority will be given to prospective PhD students.

CSE will waive the application fee to its graduate program for a limited number of applicants who attend an in-person or virtual CSE information session or recruiting event.

Visit the recruiting event page for more details and registration form.

Title: Representation Learning and Data Generation Guided by Structural and Functional Properties

Abstract: A core capability of deep learning is to learn the mapping between data and its representation. It results in two critical tasks: representation learning and data generation, which correspond to the mapping from data to representation, and the mapping from representation to data, respectively. On the other hand, in scientific and societal areas, a core goal is to explore the relation between data structure and its function, such as the molecule structure and its chemical properties, brain connectivity and mental disease, human trajectory and behavior and so on. Until now, the exploration of structure and function relation is still a largely open problem with two grand challenges: 1. Which structural properties determine a specific function? (e.g., what structural biomarkers indicate potential Alzheimer’s disease) 2. How to design a structure that has a specific function? (e.g., how to design a molecule that possesses specific properties of interests). In this talk, I will introduce our recent works on structure-guided representation learning to pursue the direction toward the first challenge and works on property-controllable data generation toward the second one. I will start with the background, and then introduce our recent work on representation learning of complex-structure data. Then I will briefly talk about our works on introducing structural inductive bias to guide the representation learning. Finally, I will talk about controllable data generation as a multi-objective optimization of the representations.

Bio: Dr. Liang Zhao is an associate professor at the Department of Computer Science at Emory University. Before that, he was an assistant professor in the Department of Information Science and Technology and the Department of Computer Science at George Mason University. He obtained his Ph.D. degree as Outstanding PhD student in 2016 from Computer Science Department at Virginia Tech in the United States. His research interests include data mining and machine learning, with special interests in spatiotemporal and network data mining, deep learning on graphs, nonconvex optimization, and interpretable machine learning. He has published over a hundred papers in top-tier conferences and journals such as KDD, TKDE, ICDM, ICLR, NeurIPS, Proceedings of the IEEE, TKDD, CSUR, IJCAI, AAAI, and WWW. He won NSF Career Award in 2020 and Jeffress Trust Award in 2019. He also won Amazon Research Award in 2020, Meta Research Award, and CRA Computing Innovation Mentor in 2021. He was ranked as “Top 20 Rising Star in Data Mining” by Microsoft Search in 2016. He won several the Best Paper Award and Candidates such as Best Paper Award in ICDM 2019, Best Paper Candidate in ICDM 2021, Best Paper Award Shortlist in WWW 2021, and Best paper Candidate in ACM SIGSPATIAL 2022. He is an IEEE senior member.

1. Richard Vuduc

2. Edmond Chow

3. Nisha Chandramoorthy

4. Nabil Imam

5. Raphael Pestourie

All the panel members have unique paths from their Ph.D. to being a professor and it's a nice opportunity for all of us to ask our questions regarding future plans after Ph.D., especially about joining academia.

Lunch will be provided at the event. Please fill out the form below to add your questions for the panel members and any dietary restrictions.

https://forms.office.com/pages/responsepage.aspx?id=u5ghSHuuJUuLem1_Mvqgg__1RjyO_H5FmaIMuCCOd_lUOEJZM0NJRTk2UE5aQlYzWUVXUElZNFk5Ui4u

Title: The Role of Applied Mathematics in the Design of Coronary Stents

Abstract: Since their introduction in the 1980s, coronary stents have undergone significant design improvements, making them a critical tool for treating severe obstructions. From original Bare-Metal Stents (BMS) to Drug Eluting Stents (DES) to the most recent experience of Bioresorbable Stents, the design of these scaffolds was minimally supported by mathematical tools. The patient-specific quantitative analysis of stented coronaries is a difficult task for the variety of complex morphologies left by the stent deployment. Therefore, this type of analysis was limited to a minimal number of patients, not compatible with clinical trials. On the other hand, the development and the failure of Bioresorbable Stents clearly pointed out the importance of rigorous quantitative tools in the design of next-generation scaffolds. In this talk, we will present recent results in investigating coronary stents based on Applied Mathematics (as opposed to traditional animal models).

We will consider in detail (1) the modeling of the elution in a multidomain problem solved by iterative substructuring methods involving simultaneously the lumen, the wall, and the struts of the stents; (2) the impact of the struts on the wall shear stress of a significant number of patients; (3) the consequent role of shape optimization and model order reduction in the design of scaffolds. 

This journey through a sophisticated combination of data and models will pinpoint the critical role of applied mathematics and scientific computing not only for a basic understanding of the biomechanics of stents, but also for the clinical routine and the design of more performing prostheses.

With the support of: NIH R01EB031101 (PI L. Brooks), NSF DMS2012286 (PI A. Veneziani) and DMS2038118 (PI: J. Nagy) 

Bio: Alessandro Veneziani is a Professor of Mathematics and Computer Science at Emory University, Atlanta, GA USA, and CSO of COVANOS, Inc. He attained his Master in Electronic Engineering (Politecnico di Milano, IT 1994) and Ph.D. in Applied Mathematics (University of Milano, IT 1998). He works on numerical modeling of cardiovascular problems, with particular attention to the translation of methods from the benchmarks to the clinics. He authored more than 90 journal papers, co-edited two research books, co-authored one textbook. He was awarded the SIAM Outstanding Paper Prize (2004 with A. Quarteroni and P. Zunino), the Giovanni Sacchi-Landriani International Prize (2007), the ACM Most Notable Paper (2014 with S. Perotto).

Realizing the envisioned system requires advancing the state-of-the-art in robotic actuation, planning, perception, and control, as well as in high-speed magnetic image acquisition and reconstruction. The presentation will focus on our work on the robotics side of the problem. I will introduce the design of the catheter and the analysis of its actuation, followed by the modeling of the catheter kinematics and optimization of the catheter design. I will present results from open-loop control experiments and dynamics response analysis of the system. I will discuss our work on robotic motion control algorithms for active cancellation of the beating heart motion, and robotic motion planning for performing different phases of intravascular cardiac interventions. Finally, I will conclude the presentation with our recent work on probabilistic active sensing algorithms that will be used to control the image acquisition system.

Bio: M. Cenk Cavusoglu is currently the Nord Professor of Engineering at the Department of Electrical, Computer, and Systems Engineering of Case Western Reserve University (CWRU), with secondary appointments in Biomedical Engineering, Mechanical and Aerospace Engineering, and Computer and Data Sciences. He received the B.S. degree in Electrical and Electronic Engineering from the Middle East Technical University, Ankara, Turkey, in 1995, and the M.S. and Ph.D. degrees in Electrical Engineering and Computer Sciences from the University of California, Berkeley, in 1997 and 2000, respectively. He was a Visiting Researcher at the INRIA Rhones-Alpes Research Center, Grenoble, France (1998); a Postdoctoral Researcher and Lecturer at the University of California, Berkeley (2000-2002); and, a Visiting Associate Professor at Bilkent University, Ankara, Turkey (2009-2010).

Dr. Cavusoglu's research spans the general areas of robotics and human-machine interfaces with special emphasis on medical robotics, and haptics. Specifically, for the past 25+ years, he has been conducting research on all of the different aspects of medical robotic systems from control, mechanism, and system design, to human-machine interfaces, haptics, and algorithms.

More information on Dr. Cavusoglu's research can be found at his homepage at
https://engineering.case.edu/cenkcavusoglu

BIO: Fuxin Li is currently an associate professor in the School of Electrical Engineering and Computer Science at Oregon State University. He has held research positions in Apple Inc., University of Bonn and Georgia Institute of Technology. He had obtained a Ph.D. degree in the Institute of Automation, Chinese Academy of Sciences in 2009. He has won an NSF CAREER award, an Amazon Research Award, (co-)won the PASCAL VOC semantic segmentation challenges from 2009-2012, and led a team to the 4th place finish in the DAVIS Video Segmentation challenge 2017. He has published more than 70 papers in computer vision, machine learning and natural language processing. His main research interests are point cloud deep networks, human understanding of deep learning, video object segmentation, multi-target tracking and uncertainty estimation in deep learning.

Bio: Aaron M. Johnson is an Associate Professor of Mechanical Engineering at Carnegie Mellon University, with additional appointments in the Robotics Institute and Electrical & Computer Engineering departments. He received his PhD from the University of Pennsylvania in 2014. His research interests are in hybrid systems, state estimation, control, legged robots, and field robotics. He is the recipient of the NSF Career award, the ARO Young Investigator Award, and the Best Paper award at the ICRA Workshop on Legged Robots, among other awards.

Preregistration through the link below is required to attend session. Registration is required at least 24 hours in advance. Zoom link will be sent to registered email addresses prior to information session.

Register here through this form

Preregistration through the link below is required to attend session. Registration is required at least 24 hours in advance. Zoom link will be sent to registered email addresses prior to information session.

Register here through this form

The symposium is a chance for faculty to meet new robotics students on campus, as well as a chance to get a better idea of what IRIM colleagues are up to these days. The goal of the symposium is to spark new ideas, new collaborations, and even new friends!

9:00am  | Registration

9:30am | Seth Hutchinson - Welcome and Introductory Remarks
9:45am | Sehoon Ha - Learning Interactive Skills of Legged Robots
10:00am | Stephen Balakirsky - Knowledge Driven Robotics
10:15am | Matthew Gombolay - RoboFederer: Towards Human-Robot Team Athletics
10:30am | Sonia Chernova - Autonomy in Human Spaces

10:45am | Break

11:00am | Danfei Xu - Towards a Data Flywheel for Long-horizon Manipulation
11:15am | Harish Ravichandar - New Wine in an Old Bottle: Classical Methods as Scaffolding for Robot Learning
11:30am | Zsolt Kira - Vision and Language Foundation Models for Open-World Robotics
11:45am | James Hayes - Understanding Grasping, Contact, and Pressure
12:00pm | Animesh Garg - Science and Systems for Foundation Models Manipulation

12:15pm - 1:15pm | Lunch [Box lunches provided to registered attendees]

1:15pm | Ai-Ping Hu - Agricultural Robotics at Georgia Tech
1:30pm | W. Hong Yeo - Smart and Connected Bioelectronics for Integration with Robots
1:45pm | Ellen Yi Chen Mazumdar - Soft and Additively Manufacturable Electromagnetic Actuators
2:00pm | Yue Chen - Tentacle-like Continuum Robots
2:15pm | Greg Sawicki - Beyond reducing cost of transport: What can lower-limb exoskeletons do on the shortest and longest timescales?
2:30pm | Ye Zhao- Can humanoid robots move and manipulate like humans?

2:45pm | Break

3:00pm | Spyros Reveliotis - Traffic Control of Multi-Agent Systems Circulating on Graphs
3:15pm | Shreyas Kousik - Towards Safe Full-Stack Design for Autonomous Robots
3:30pm | Charlie Kemp - TBA
3:45pm | Panos Tsiotras – Large-scale Stochastic Team Games

4:15pm | Closing Remarks

4:30pm - 6:30pm | Poster Session

6:00pm - 9:00pm | Robotics Meet-Up

Register Here

Where: Klaus Atrium
266 Ferst Drive NW
When: Thursday, April 20, 11:30 a.m. - 1 p.m.

(registration required)

Date: Tuesday, April 4, 2023 at 11:00 am

Location: Coda 114

Link: This seminar is an in-person event only. However, the seminar will be recorded and uploaded to the School of Computational Science and Engineering channel on Georgia Tech MediaSpace following the presentation.

Title: Learning from the Interconnected World with Graphs

Abstract: The fact that our world is fundamentally interconnected presents unique challenges for modern data-driven research. In this talk, I will present my research on investigating the interconnected world through the lens of graphs. Specifically, I will present my research on accelerating graph AI research by systematically investigating the design space and task space for graph deep learning. The research has democratized graph AI for domain experts and helped them with scientific discoveries. Next, I will demonstrate my pioneering research in deep graph generative models which can generate novel realistic graph structures toward desirable objectives. This line of work has broad applications in molecule design and drug discovery. Lastly, I will show that graphs can further power AI problems in general domains. I will specifically cover my research in representing neural networks as relational graphs, which advances the design and understanding of deep neural networks and connects to network science and neuroscience. Overall, the talk will outline the promising path toward bridging interdisciplinary research and extending the frontiers of AI with graphs.

Bio: Jiaxuan You received his CS Ph.D. from Stanford University, advised by Prof. Jure Leskovec. His research investigates scientific and industrial problems through the lens of graphs and develops graph AI methods to solve these problems. He has published 12 first-author papers in NeurIPS, ICML, ICLR, AAAI, KDD, and WWW, many of which are widely recognized. He has created or co-led multiple open-source software with over 20,000 combined GitHub stars. Jiaxuan has received multiple prestigious awards, including a JPMorgan Chase Ph.D. Fellowship, AAAI Best Student Paper Award, World Bank Best Big Data Solution, and Outstanding TA Award from Stanford CS. He was the lead organizer of NeurIPS New Frontiers in Graph Learning Workshop, a co-organizer of the Stanford Graph Learning Workshop, and a program committee member at numerous top-tier AI conferences and journals. His Ph.D. research further leads to a startup Kumo AI which demonstrates significant real-world impact. Webpage: https://cs.stanford.edu/~jiaxuan.

Alyssa Sheehan, Designing for the Future of Meaningful Work
My research is focused on evaluating emerging workplace technologies to shape the Future of Work. In this talk, I will share one example of how tools like augmented reality implemented on the manufacturing shop floor change the nature of work for blue-collar professionals. My work points to the need for computing design to address different value systems across blue- and white-collar domains. Specifically, we need to consider the professional boundaries these tools remake to create technologies that positively impact the future of work – not just developing tools that are useful but creating jobs that are meaningful in the face of automation and the enabling data economies. Designing for meaningful work provides a lens that prioritizes individual worker values and attends to existing organizational norms necessary for the design and use of effective automated systems.

Vedant Das Swain, Information Workers Perspectives on Phenotyping Performance and Wellbeing with Passive Sensing Enabled AI
We live in a time when our conception of a thriving worker is in flux. These changing definitions significantly affect information workers, who are increasingly unsatisfied with the care they get at work. To provide workers insight into successful behaviors, my research has explored the potential of passive sensing to algorithmically estimate performance and wellbeing. I refer to these approaches as Passive Sensing-enabled AI (PSAI). While hybrid work paradigms have simultaneously created new opportunities for PSAI, but have also fostered anxieties of misuse and privacy intrusions within a power asymmetry. At this juncture, it is unclear if those who are sensed can find these systems acceptable. In this talk, I will first introduce the state-of-the-art PSAI for information workers. Then, I will share my findings from scenario-based interviews of 28 information workers to highlight their perspectives as data subjects in PSAI. I unpack their expectations using the Contextual Integrity framework of privacy and information gathering. Participants described the appropriateness of PSAI based on its impact on job consequences, work-life boundaries, and preservation of flexibility. They also perceived that PSAI inferences could be shared with selected stakeholders if they could negotiate the algorithmic inferences. My research takes a step towards worker-centric approaches to implementing PSAI as an empowering tool in the future of work.

Yixuan ("Janice") Zhang, The Rise & Fall of Online Trust
Online trust is becoming a central issue in modern society. Trusting fallacies can harm the wellbeing of individuals and society as a whole, especially during crises like COVID-19. During crises, the inconclusive, conflicting, and time-sensitive information created by varied sources and disseminated on diverse platforms can engender a sense of confusion and uncertainty. This uncertainty makes it difficult for people to determine what and whom they should trust. In this talk, I will share several empirical studies to answer: how people form and develop their online trust over time and how content producers (human & AI) approach “trustworthy” design. I will then discuss ongoing and future efforts in building tools to address issues of online trust, investigating the risks of AI to prevent “manipulating” trust, and developing theoretical frameworks to guide rigorous future research.

Speaker Bios:

Alyssa Sheehan is currently a PhD candidate in Human Centered Computing at Georgia Institute of Technology and the Director for the Georgia Center of Innovation for Aerospace. Her industry experience working in manufacturing has directly informed her research which is focused on evaluating and designing automation for the future of work in blue collar domains. Together with industry partners, Alyssa investigates the impact of new technologies on workplace practices in settings including manufacturing, logistics, and emergency response to inform technology design and data practices. Situated within the field of Human Computer Interaction (HCI), her research has been published in top tier computing venues including Association for Computing Machinery Computer Human Interaction, Computer Supported Cooperative Work, and Designing Interactive Systems.

Vedant Das Swain is a PhD Candidate in the School of Interactive Computing at the Georgia Institute of Technology, advised by Munmun De Choudhury and Gregory Abowd. His research contributes to the future of work and behavioral wellbeing in general. He identifies, develops, and critiques opportunities to leverage ubiquitous technologies for algorithmic inference of performance and mental wellbeing. He consistently works with organizational psychologists to inform his investigations and also collaborates with Microsoft Research to develop better tools for worker wellbeing. His research has been published at top-tier computing venues like CHI, CSCW, UbiComp/IMWUT, ACII, and IEEE CogMI. His paper at CHI 2022 won a Best Paper Honorable Mention award. He is the winner of the Gaetano Borriello Outstanding Student Award at UbiComp 2022 and the GVU Foley Scholar Award 2022. His research has been supported by IARPA, NSF, CDC, ORNL, and Semiconductor Research Corporation.

Yixuan ("Janice") Zhang is a PhD candidate in Human-Centered Computing at Georgia Tech. Her research broadly lies in human-computer interaction (HCI), data visualization (VIS), equitable computing, and the interdisciplinary fields of crisis informatics and health informatics. She regularly publishes in premier HCI and visualization venues, such as ACM CHI, CSCW, and IEEE VIS. Her research has been well recognized by academia: she was named a Rising Star in EECS 2022, and selected as a Foley Scholar. The impact of her research extends beyond academia; for example, she was invited to present her work to broader audiences, such as the World Health Organization (WHO). More information is available at https://zjanice.github.io

How to watch: If you can't attend. please watch the Live Stream, or view the Recording (available 30 days after event). If you have any questions, email us at gvu@cc.gatech.edu.

Schedule of Brown Bag Speakers Spring 2023

There is a growing global consensus that Africa is the youngest and fastest-growing continent in the world. By 2030, it is estimated that over 300 million young people will seek employment in Africa. This demographic boom will push Africa’s workforce to more than a billion people, the largest in the world. Youth employment is key to building prosperity in South Africa. The changing nature of work means that digital skills are at the heart of the question of how to create sustainable youth livelihoods in South Africa. Building digital skills is a daunting task, as it is near impossible to keep up with the pace of change. Indeed, the digital revolution has overthrown the very way that we learn. How, then, do we think about what digital skills to impart, and how best to impart them to secure positive results for the earning potential of young women?

Unemployment in South Africa currently sits at over 29%, the highest in a decade. According to a report published by McKinsey on ‘The future of work in South Africa: Digitisation, productivity and job creation’ in 2019, digitization could result in a net gain of more than 1m jobs by 2030 and could create 1.6m jobs for women – and boost empowerment. Tech-enabled jobs will require higher skills, resulting in demand for an additional 1.7m graduates. To seize the opportunity, action is needed by government, business and individuals. To navigate these labor-market transitions successfully, women will need different skills and more education, mobility to switch jobs easily, and access to technological capabilities that will not only be in demand but can also open up new ways of working and new sources of economic opportunity. Women face persistent challenges on these three dimensions that will be needed to thrive in the automation era; these long-established structural and societal challenges have already slowed women’s progress toward gender equality in work. Digital and internet technologies offer women a way to break down barriers by making reskilling more accessible and enabling flexible working, for instance. Moreover, private and public sector leaders have a huge opportunity to support women as they navigate impending transitions.

A mismatch exists between skills gained at the universities and the requirements of the job market due to the disconnect between the key stakeholders – employers, government, academia, and the students themselves. Millions are under- or unemployed, yet nearly two-thirds of companies report having positions for which they cannot find qualified applicants. Meanwhile, as technology increasingly reshapes the future of work, more than 50% of today’s jobs require highly specialized technical skills, and 77% will require them in less than a decade. In this context, women are most affected.

Intend to address three priorities:

1. To increase access to digital skills, knowledge and use of emerging technologies for young women in South Africa;
2. To create or strengthen effective links between digital skills development and labor market demand;
3. To establish or enable synergies in the skills and labor ecosystem, including with the private sector; facilitate conditions for on-the-job skills transfer.

Conclusion:

1. Economic opportunities of women are improved thanks to the use of digital technologies;
2. Local digital skills, innovation and technology transfer are stimulated to support the development of economic opportunities for women;
3. Enabling ecosystem for inclusive innovation and knowledge transfer is developed.

Research will include:

• Data collection efforts to measure social mobility and equal opportunity, and understand the drivers of these issues;
• Discussion of how policies can break down barriers to equal opportunity and promote social mobility;
• Analysis of the role of civil society and the private sector in fostering equal opportunity.

Speaker Bio:

Mbali Hlongwane is a systems engineer, founder and leader, project lead, ecosystem manager, and public speaker. "My personal vision is to drive the economic, innovative, and social development of Africa at large through developing and positioning women to participate in the tech transformation. I believe that you simply can't run a good company or a good country without a woman being involved. Due to the existing gap within the tech sector, I want to ensure that 1 million women in Africa have tech skills to build solutions, start companies, and take up executive roles in tech organizations by 2030 driven by technology, knowledge sharing, integrity, and sustainability."

How to watch: If you can't attend. please watch the Live Stream, or view the Recording (available 30 days after event). If you have any questions, email us at gvu@cc.gatech.edu.

Schedule of Brown Bag Speakers Spring 2023

There is a growing global consensus that Africa is the youngest and fastest-growing continent in the world. By 2030, it is estimated that over 300 million young people will seek employment in Africa. This demographic boom will push Africa’s workforce to more than a billion people, the largest in the world. Youth employment is key to building prosperity in South Africa. The changing nature of work means that digital skills are at the heart of the question of how to create sustainable youth livelihoods in South Africa. Building digital skills is a daunting task, as it is near impossible to keep up with the pace of change. Indeed, the digital revolution has overthrown the very way that we learn. How, then, do we think about what digital skills to impart, and how best to impart them to secure positive results for the earning potential of young women?

Unemployment in South Africa currently sits at over 29%, the highest in a decade. According to a report published by McKinsey on ‘The future of work in South Africa: Digitisation, productivity and job creation’ in 2019, digitization could result in a net gain of more than 1m jobs by 2030 and could create 1.6m jobs for women – and boost empowerment. Tech-enabled jobs will require higher skills, resulting in demand for an additional 1.7m graduates. To seize the opportunity, action is needed by government, business and individuals. To navigate these labor-market transitions successfully, women will need different skills and more education, mobility to switch jobs easily, and access to technological capabilities that will not only be in demand but can also open up new ways of working and new sources of economic opportunity. Women face persistent challenges on these three dimensions that will be needed to thrive in the automation era; these long-established structural and societal challenges have already slowed women’s progress toward gender equality in work. Digital and internet technologies offer women a way to break down barriers by making reskilling more accessible and enabling flexible working, for instance. Moreover, private and public sector leaders have a huge opportunity to support women as they navigate impending transitions.

A mismatch exists between skills gained at the universities and the requirements of the job market due to the disconnect between the key stakeholders – employers, government, academia, and the students themselves. Millions are under- or unemployed, yet nearly two-thirds of companies report having positions for which they cannot find qualified applicants. Meanwhile, as technology increasingly reshapes the future of work, more than 50% of today’s jobs require highly specialized technical skills, and 77% will require them in less than a decade. In this context, women are most affected.

Intend to address three priorities:

1. To increase access to digital skills, knowledge and use of emerging technologies for young women in South Africa;
2. To create or strengthen effective links between digital skills development and labor market demand;
3. To establish or enable synergies in the skills and labor ecosystem, including with the private sector; facilitate conditions for on-the-job skills transfer.

Conclusion:

1. Economic opportunities of women are improved thanks to the use of digital technologies;
2. Local digital skills, innovation and technology transfer are stimulated to support the development of economic opportunities for women;
3. Enabling ecosystem for inclusive innovation and knowledge transfer is developed.

Research will include:

• Data collection efforts to measure social mobility and equal opportunity, and understand the drivers of these issues;
• Discussion of how policies can break down barriers to equal opportunity and promote social mobility;
• Analysis of the role of civil society and the private sector in fostering equal opportunity.

Speaker Bio:

Mbali Hlongwane is a systems engineer, founder and leader, project lead, ecosystem manager, and public speaker. "My personal vision is to drive the economic, innovative, and social development of Africa at large through developing and positioning women to participate in the tech transformation. I believe that you simply can't run a good company or a good country without a woman being involved. Due to the existing gap within the tech sector, I want to ensure that 1 million women in Africa have tech skills to build solutions, start companies, and take up executive roles in tech organizations by 2030 driven by technology, knowledge sharing, integrity, and sustainability."

How to watch: If you can't attend. please watch the Live Stream, or view the Recording (available 30 days after event). If you have any questions, email us at gvu@cc.gatech.edu.

Schedule of Brown Bag Speakers Spring 2023

This talk will address the design challenges and opportunities in creating health technologies for men to improve the health of women in religiously conservative, patriarchal, and low-income societies. In this talk, I will share findings from the deployment of a speech-based service called Super Abbu (Super Dad) designed to connect expectant fathers to doctors and to each other. Over a period of 71 days, the service reached upwards of 20,000 users who spent almost 400 thousand minutes on the platform. Through a critical examination of cultural and societal factors, such as traditional gender roles, stigma towards sexual health information-seeking, and limited access to resources, I will highlight key considerations for designing effective and culturally sensitive health technologies for this population. The goal of this talk is to provide insights and recommendations for designers, researchers, and practitioners to create health technologies that are inclusive, accessible, and effective for users, regardless of their cultural, social, and economic backgrounds.

Speaker Bio:

Mustafa Naseem is a Clinical Assistant Professor at the University of Michigan School of Information. Mustafa’s research focuses on design challenges around creating, deploying and scaling health technologies in religiously conservative, patriarchal and low-income societies. Mustafa’s scholarship informs technology and policy design in international development, with a focus on centering marginalized individuals’ voices and needs, while focusing on factors such as access and gender. Mustafa has taught and conducted research with underserved communities in Pakistan, Ghana, South Africa, Tanzania, Botswana and the US. Mustafa’s work is published in venues including ACM CHI, CSCW, Web Conference (WWW), Compass, and ICTD, and has received Best Paper Honorable Mention and Diversity and Inclusion Awards at ACM CHI and CSCW. Mustafa received his Masters in ICT and Development (ICTD) from the University of Colorado Boulder on a Fulbright Scholarship. His work is supported by the National Science Foundation, the National Institutes of Health, US National Academy of Sciences, Engineering, and Medicine, Bill and Melinda Gates Foundation, and USAID, among others.

How to watch: If you can't attend. please watch the Live Stream, or view the Recording (available 30 days after event). If you have any questions, email us at gvu@cc.gatech.edu.

Schedule of Brown Bag Speakers Spring 2023

“Deepfakes” are videos in which the (usually human) subject of a video has been digitally altered to appear to do or say something that they never actually did or said. Sometimes these manipulations produce innocuous novelties (e.g., testing what it would look like if Will Smith had been cast as “Neo” in the film The Matrix), but far more dangerous use cases have been observed (e.g., producing fake footage of Ukrainian President, Volodymyr Zelenskyy, in which he instructs Ukrainian military forces to surrender on the battlefield). Generating the knowledge and tools necessary to defend against potential harms these videos could impose is likely to rely on contributions from a broad coalition of disciplines, many of which are represented in the GVU. In this week’s Brown Bag presentation, we will offer some real-time demonstrations of deepfake technology and present findings from our work that has largely focused on investigating the psychological factors influencing deepfake detection.

Speaker Bios:

John Stasko is a Regents Professor in the School of Interactive Computing at the Georgia Institute of Technology, where he has been on the faculty since 1989. He works in the areas of information visualization and visual analytics, approaching each from a human-computer interaction perspective. Stasko received the IEEE Visualization and Graphics Technical Committee (VGTC) Visualization Technical Achievement Award in 2012 and was inducted into the ACM CHI Academy in 2016 and IEEE Visualization Academy in 2019. He was named an IEEE Fellow in 2014 and an ACM Fellow in 2023.

Richard Catrambone is a Professor in the School of Psychology at the Georgia Institute of Technology. He received his B.A. from Grinnell College and his Ph.D. in Experimental Psychology from the University of Michigan. His research interests include problem solving, educational technology, and human-computer interaction. He explores how to create instructional materials that help learners understand how to approach problems in a meaningful way rather than simply memorizing a set of steps that cannot easily be transferred to novel problems. He served on the Cognitive Science Society governing board from 2011-2016 and was chair of the Society in 2015.

Zack Tidler is a doctoral student in the School of Psychology with an emphasis in engineering psychology. His primary research interest is in developing new measures of human cognitive ability that consider tool usage, but he has developed a secondary research program which is focused on the study of deepfake video detection in humans. He is a former president of the Georgia Tech Chapter of the Human Factors and Ergonomics Society. His work on deepfake detection has been featured in the College of Sciences newsletter and the 2021 issue of the GT Alumni Magazine.

How to watch: If you can't attend. please watch the Live Stream, or view the Recording (available 30 days after event). If you have any questions, email us at gvu@cc.gatech.edu.

Schedule of Brown Bag Speakers Spring 2023

Date: Thursday, March 16, 2023 at 11:00 am

Location: Scheller College of Business, Room 102

Link: This seminar is an in-person event only. However, the seminar will be recorded and uploaded to the School of Computational Science and Engineering channel on Georgia Tech MediaSpace following the presentation.

Title: Foundation Models for Robustness to Distribution Shift

Abstract: Machine learning systems are not robust—they suffer large drops in accuracy when deployed in different environments from what they were trained on. In this talk, I show that the foundation model paradigm—adapting models that are pretrained on broad unlabeled data—is a principled solution that leads to state-of-the-art robustness. I will focus on the key ingredients: how we should pretrain and adapt models for robustness. (1) First, I show that contrastive pretraining on unlabeled data learns transferable representations that improves accuracy even on domains where we had no labels. We explain why pretraining works in a very different way from some classical intuitions of collapsing representations (domain invariance). Our theory predicts phenomena on real datasets, and leads to improved methods. (1) Next, I will show that the standard approach of adaptation (updating all the model's parameters) can distort pretrained representations and perform poorly out-of-distribution. Our theoretical analysis leads to better methods for adaptation and state-of-the-art accuracies on ImageNet and in applications such as satellite remote sensing, wildlife conservation, and radiology.

Bio: Ananya Kumar is a Ph.D. candidate in the Department of Computer Science at Stanford University, advised by Percy Liang and Tengyu Ma. His work focuses on representation learning, foundation models, and reliable machine learning. His papers have been recognized with several Spotlight and Oral presentations at NeurIPS, ICML, and ICLR, and his research is supported by a Stanford Graduate Fellowship.

Date: Tuesday, March 14, 2023 at 11:00 am

Location: Scheller College of Business, Room 101

Link: This seminar is an in-person event only. However, the seminar will be recorded and uploaded to the School of Computational Science and Engineering channel on Georgia Tech MediaSpace following the presentation.

Title: Towards Data-Driven Modeling in Large-Scale Naturalistic Neuroscience

Abstract: Neuroscience is currently undergoing an explosion in the availability of large-scale brain activity data, so the major challenge no longer lies in data collection, but in deriving understanding from this abundant stream of complex, high-dimensional, noisy data with methods that fully leverage its potential. How can we understand neural representations and infer computational principles from large-scale brain activity data directly? My research group will develop the theory, modeling, and machine learning techniques needed to tackle these challenges, and will provide algorithmically and computationally precise accounts of information processing in the brain. Progress in this research could advance our understanding of biological intelligence and the neural basis of neuropsychiatric disorders, inform novel treatments and neural prostheses, and lead to the development of new approaches to generating machine intelligence.

In this talk, I will present lines of previous and proposed research that highlight the potential of this vision. First, I will present a line of data-driven modeling that revealed the representational structure in the high-level visual cortex and led to the discovery of a neural population selectively responsive to images of food. Second, I will present a modeling framework, called response-optimization, for inferring computations directly from brain activity data with minimal apriori hypotheses. Here, we trained artificial neural network (ANN) models directly to predict the brain activity related to viewing natural images. We then developed techniques for interpreting the networks and characterizing the emergent functional capabilities of these brain response-optimized networks. This work highlights how models trained to capture human brain activity can spontaneously recapitulate human-like behavior. Third, I will present my work on developing neural network models of brain responses across wide-spread cortical regions to dynamic, multi-modal stimuli like movies, with an integrated modeling approach that captured visual attention, multi-sensory auditory-visual interactions and temporal context. Finally, I will propose future directions for leveraging these data-driven computational tools toward i) understanding the representational structure underlying our capacity for high-level cognitive processes like social perception, ii) understanding the granularity of similarity between biological and artificial neural networks, iii) understanding the functional consequences of brain alignment in ANN representations, and iv) understanding not just how the brain is organized or how a process works but fundamentally, why the brain works the way it does.

Bio: Meenakshi Khosla is a postdoctoral associate in the Brain and Cognitive Sciences department at MIT. Her research interests lie at the intersection of neuroscience, artificial intelligence, and large-scale data analysis. Her current projects focus on leveraging large-scale data to develop interpretable machine learning tools for understanding structured neural representations and computations in biological systems.  Previously, Meenakshi received her PhD in Electrical and Computer Engineering (ECE) from Cornell University, where she worked broadly at the intersection of machine learning and neuroimaging, developing predictive models to understand the distinctive characteristics of the brains of people affected with different mental disorders.

This talk will address the design challenges and opportunities in creating health technologies for men to improve the health of women in religiously conservative, patriarchal, and low-income societies. In this talk, I will share findings from the deployment of a speech-based service called Super Abbu (Super Dad) designed to connect expectant fathers to doctors and to each other. Over a period of 71 days, the service reached upwards of 20,000 users who spent almost 400 thousand minutes on the platform. Through a critical examination of cultural and societal factors, such as traditional gender roles, stigma towards sexual health information-seeking, and limited access to resources, I will highlight key considerations for designing effective and culturally sensitive health technologies for this population. The goal of this talk is to provide insights and recommendations for designers, researchers, and practitioners to create health technologies that are inclusive, accessible, and effective for users, regardless of their cultural, social, and economic backgrounds.

Speaker Bio:

Mustafa Naseem is a Clinical Assistant Professor at the University of Michigan School of Information. Mustafa’s research focuses on design challenges around creating, deploying and scaling health technologies in religiously conservative, patriarchal and low-income societies. Mustafa’s scholarship informs technology and policy design in international development, with a focus on centering marginalized individuals’ voices and needs, while focusing on factors such as access and gender. Mustafa has taught and conducted research with underserved communities in Pakistan, Ghana, South Africa, Tanzania, Botswana and the US. Mustafa’s work is published in venues including ACM CHI, CSCW, Web Conference (WWW), Compass, and ICTD, and has received Best Paper Honorable Mention and Diversity and Inclusion Awards at ACM CHI and CSCW. Mustafa received his Masters in ICT and Development (ICTD) from the University of Colorado Boulder on a Fulbright Scholarship. His work is supported by the National Science Foundation, the National Institutes of Health, US National Academy of Sciences, Engineering, and Medicine, Bill and Melinda Gates Foundation, and USAID, among others.

How to watch: If you can't attend. please watch the Live Stream, or view the Recording (available 30 days after event). If you have any questions, email us at gvu@cc.gatech.edu.

Schedule of Brown Bag Speakers Spring 2023