Title: 

Amir Yazdanbakhsh (Google DeepMind), Suvinay Subramanian (Google)

Teams Link

Abstract:

The rapid advancement of artificial intelligence (AI) has ushered in an era of unprecedented computational demands, necessitating continuous innovation in computing systems. In this talk, we will highlight how codesign has been a key paradigm in enabling innovative solutions and state-of-the-art performance in Google's AI computing systems, namely Tensor Processing Units (TPUs). We present several codesign case studies across different layers of the stack, spanning hardware, systems, software, algorithms, all the way up to the datacenter. We discuss how TPUs have made judicious, yet opinionated bets in our design choices, and how these design choices have not only kept pace with the blistering rate of change, but also enabled many of the breakthroughs in AI.

Bio:

Amir Yazdanbakhsh received his Ph.D. degree in computer science from the Georgia Institute of Technology. His Ph.D. work has been recognized by various awards, including Microsoft PhD Fellowship and Qualcomm Innovation Fellowship. Amir is currently a Research Scientist at Google DeepMind where he is the co-founder and co-lead of the Machine Learning for Computer Architecture team. His work focuses on leveraging the recent machine learning methods and advancements to innovate and design better hardware accelerators. He is also interested in designing large-scale distributed systems for training machine learning applications, and led the development of a massively large-scale distributed reinforcement learning system that scales to TPU Pod and efficiently manages thousands of actors to solve complex, real-world tasks. The work of our team has been covered by media outlets, including WIRED, ZDNet, AnalyticsInsight, InfoQ. Amir was inducted into the ISCA Hall of Fame in 2023.

Suvinay Subramanian is a Staff Software Engineer at Google, where he works on the architecture and codesign for Google's ML supercomputers, Tensor Processing Units (TPUs). His work has directly impacted innovative architecture and systems features in multiple generations of TPUs, and empowered performant training and serving of Google's research and production AI workloads. Suvinay received a Ph.D. from MIT, and a B.Tech from the Indian Institute of Technology Madras. He also co-hosts the Computer Architecture Podcast that spotlights cutting-edge developments in computer architecture and systems.

Abstract:
Quantum computing presents many challenges for the programming language community. How can we program quantum algorithms in a way that ensures they behave correctly? In this talk, I will discuss how types can be used to enforce various properties of quantum programs. I will first talk about how linear types and dependent types can be useful for programming quantum circuits. I will then discuss my recent work on designing a type system to enable the interaction of quantum circuit generation time and quantum circuit execution time. If time permits, I will sketch how to ensure reversibility and controllability of the quantum circuits using types.

Bio:
Frank (Peng) Fu is an assistant professor in the Computer Science and Engineering Department at the University of South Carolina. Previously, he was a postdoctoral researcher at Dalhousie University in Canada. He obtained his Ph.D. degree from University of Iowa. His research interests are in quantum programming languages, type theory and their applications.

Location:

In-person

Innovation Center Building 1400

Virtual audience
 

Prof. Mahanth Gowda from the Pennsylvania State University.

Teams Meeting link

See https://acm.cse.sc.edu/events/2023/2023-09-28/

SUMMARY: This talk will present an overview of recent research at UW FUNLab around the use of vehicular radar for advanced driver assistance systems (en route to a future vision of autonomous driving). Wideband (typically FMCW or chirp) radars are increasingly deployed onboard vehicles as key high-resolution sensors for environmental mapping or imaging and various safety features. The talk will be demarcated into two parts, centered on the evolving role of radar ‘cognition’ in complex operating environments to address two important future challenges:
 

1. Mitigating multi-access interference among Radars (e.g., dense traffic scenario)
   This will first illustrate the impact of mutual interference on detection performance in commercial Chirp/FMCW radars and then highlight some multi-access protocol design approaches for effective resource sharing among multiple radars.
2. Contributions to radar vision via new radar hardware (MIMO radar) + associated advanced signal processing (Synthetic Aperture) principles using Convolutional Neural Network (‘Radar Net’) based machine learning approaches for enhanced object detection/classification in challenging circumstances.

Abstract

Large Language Models (LLMs) have garnered significant attention from researchers, including clinicians, due to their ability to respond to various human queries. Innovations like ChatGPT's groundbreaking reinforcement learning with human feedback and Google's domain-specific fine-tuning in Med-PaLM have introduced two potent information-providing platforms for general health inquiries. The 2023 Gartner Hype Curve places such LLMs at the pinnacle, foreseeing translational impact in the next 2-3 years. This foresight is grounded in comprehensive assessments of recent studies that have illuminated the limitations of these LLMs.

The remarkable potential of these LLMs, when fortified with features like human-level explainability, consistency, reliability, and safety, holds the promise of making deployable systems usable and readily adaptable to various scenarios where human lives may be affected. The talk will introduce a suite of methodologies (methods+metrics) under the Knowledge-powered CREST Framework for LLMs. This practical approach harnesses declarative, procedural, and graph-based knowledge within a neurosymbolic framework to shed light on the challenges associated with LLMs. 
 

Bio

Manas Gaur is an assistant professor in the Department of Computer Science and Electrical Engineering at the University of Maryland, Baltimore County (UMBC). At UMBC, he leads the Knowledge-infused AI and Inference (KAI2) lab. Before entering academia, he was the lead research scientist in Natural Language Processing (NLP) at the AI Center within Samsung Research America. He also held a visiting researcher role at the Alan Turing Institute. Dr. Gaur earned his Ph.D. under the guidance of Prof. Amit P. Sheth at the Artificial Intelligence Institute, University of South Carolina. Together, they played a pivotal role in the development of Knowledge-infused Learning, a paradigm that harmonizes seamlessly with NeuroSymbolic AI. He has been recognized as AAAI New Faculty for 2023 and is currently an advisor to Balm.ai, a startup on Mental Health. More details about him are at: https://manasgaur.github.io/
 

Location:

In-person

Innovation Center Building 1400

Online

Abstract
The “Curse of Dimensionality” issue of dynamic programming-based control approaches for large-scale state and action space of dynamic systems or agents led to the development of approximate dynamic programming (ADP). The approximate dynamic programming unifies the theory of optimal control, adaptive control, and reinforcement learning (RL) to obtain an approximate solution to the Bellman equation online and forward-in-time. In general, the value function, which is the solution to the Bellman equation in a discrete-time framework or Hamilton-Jacobi-Equation (HJB) in a continuous-time framework, is approximated using a neural network-based approximator.  The learning/adaptive nature of the solution often partially or fully relaxes the assumption of complete system information, which leads to optimal decision-making in uncertain/unknown environments. This presentation will traverse the evolution of the ADP/RL-based optimal control designs for dynamic cyber-physical systems, moving from traditional iterative solutions to those that emphasize time-based solutions. Specifically, there will be a focus on computation and communication-saving aspects of the ADP/RL-based designs.  The resource-aware ADP scheme, referred to as event-driven ADP, using Q-learning and Temporal Difference learning approaches will be discussed in detail. The event-driven approaches train the neural network approximators and update the actions at certain events only, thereby considerably minimizing the computational and communication requirements for the implementation of the learning-based control schemes over the communication network. Concluding this presentation, we will probe into some of the unresolved challenges of ADP/RL schemes, emphasizing their potential vulnerabilities in a cyber-physical framework.

Bio
Avimanyu Sahoo received his Ph.D. in Electrical Engineering from  Missouri University of Science and Technology, Rolla, MO, USA, in 2015 and a Masters of Technology (MTech) and the Indian Institute of Technology (BHU), Varanasi, India, in 2011. He is currently an Assistant Professor in the Electrical and Computer Engineering Department at the University of Alabama in Huntsville (UAH), AL. Prior to joining UAH, Dr. Sahoo was an Associate Professor in the Division of Engineering Technology at Oklahoma State University, Stillwater, OK.

Dr. Sahoo’s research interest includes learning-based control and its applications in lithium-ion battery pack modeling, diagnostics, prognostics, cyber-physical systems, and electric machinery health monitoring. Currently, his research focuses on developing intelligent battery management systems (BMS) for lithium-ion battery packs used onboard electric vehicles, computation, and communication-efficient distributed intelligent control schemes for cyber-physical systems using approximate dynamic programming, reinforcement learning, and distributed adaptive state estimation.

Link
 

Abstract

Ever since ChatGPT was launched last November, it has captured the public's imagination quickly leading to upheaval and excitement in all communities, whether business, research or government. But how is it changing the AI research community from where the technology came about? In this talk, I will take a practical perspective on the potential and challenges of working with Large Language Model (LLM)-based technologies. We are using it for core AI tasks of generating plans and knowledge graphs, and exploring its use for decision support in finance, water and elections. The TL;DR is that LLMs can be quite useful but unreliable, and this opens up exciting research opportunities for trusted AI.

Bio

Biplav Srivastava is a Professor of Computer Science at the AI Institute and Department of Computer Science at the University of South Carolina which he joined in 2020 after two decades in industrial research. He directs the 'AI for Society' group which is investigating how to enable people to make rational decisions despite the real world complexities of poor data, changing  goals and limited resources by augmenting their cognitive limitations with technology. His work in Artificial Intelligence spans the sub-fields of reasoning (planning, scheduling), knowledge extraction and representation (ontology, open data), learning (classification, deep, adversarial) and interaction (collaborative assistants), and extends to their application for Services (process automation, composition) and Sustainability (water, traffic, health, governance). In particular, he has been involved with building innovative systems for decision support in domains as diverse as governance (IJCAI 2016), astronomy (AAAI 2018 best demo award), water (AAAI 2018), smart room (ICAPS 2018 demo runner up, IJCAI 2018), career planning (commercial product), market intelligence (AAAI 2020 deployed AI award), dialogs for information retrieval (ICAPS 2021), fairness assessment (AAAI 2021),  computer games (AAAI 2022), generalized planning (IJCAI 2023), transportation, set recommendation (teaming, meals) and health. Biplav’s works have led to many science firsts and high-impact commercial innovations valued over billions of dollars, 200+ papers and 70 US patents issued, and awards for papers, demos and hacks. He is an ACM Distinguished Scientist, AAAI Senior Member, IEEE Senior Member and AAAS Leshner Fellow for Public Engagement on AI (2020-2021). More details about him are at:

https://sites.google.com/site/biplavsrivastava/

Zoom Link

DISSERTATION DEFENSE 
Author : William Hoskins

Advisor : Dr. Jijun Tang

Date : August 7, 2023

Time: 12:30 pm - 1:30 pm

Place : Virtual

Meeting Link:

Abstract 

In this work we discuss the design and development of the Carolina Automated Reading Evaluation (CARE), created to facilitate the finding of deficits in the reading ability of children from four to nine years of age. Designed to automate the process of screening for reading deficits, the CARE is an interactive computer-based tool that helps eliminate the need for one-on-one evaluations of pupils to detect dyslexia and other reading deficits and facilitates the creation of new reading tests within the platform.  

While other tests collect specific data points in order to determine whether a pupil has dyslexia, they typically focus on only a few metrics for diagnosis, such as handwriting analysis or eye tracking. The CARE collects data across up to 16 different subtests, each built to test proficiency in various reading skills. These skills include reading fluency, phoneme manipulation, sound blending, and many other essential skills for reading. This wide variety of measurements allows for a more focused intervention to be created for the pupil. 

The first chapter of this work recounts the design and development process for the CARE platform, describing the creation of the test development tools and the individual subtests. The second chapter focuses on using eye tracking to optimize the teacher facing user interface. Chapter three discusses the usage of reinforcement learning to create a Computerized Adaptive Test for the CARE.