Prof. Tyler Derr
Vanderbilt University

Data-Centric AI for Real-World Graph Applications

In today's data-driven world, graphs play a crucial role in representing a significant portion of big data across diverse domains. However, real-world graph data often presents challenges in terms of quality. This talk emphasizes the need for data-centric graph learning, which has primarily been overlooked in favor of model-centric approaches. We showcase recent advancements in recommender systems, neuroimaging, and drug discovery, underscoring the critical role of prioritizing data-centric approaches to address data quality challenges across varied domains. By tackling data quality issues and optimizing graph datasets, we unlock the potential for substantial performance enhancements in real-world applications employing graph neural networks. The talk concludes by offering insights into future directions and potential advancements, fostering a deeper understanding of the evolving landscape in this field.

Biography

Dr. Tyler Derr is an Assistant Professor in the Department of Computer Science at Vanderbilt University. He directs the Network and Data Science (NDS) lab, which conducts research in the areas of data mining and machine learning, with emphasis on social network analysis, deep learning on graphs, and responsible AI for social good. He is actively involved in his field, both in terms of publishing and serving on international conference organizing committees, e.g., KDD and WSDM. He is also an Associate Editor for related journals including IEEE Transactions on Big Data along with Tsinghua Science and Technology. He has received numerous prestigious awards, such as the Teaching Innovation Award from Vanderbilt's School of Engineering and the National Science Foundation CAREER Award. More details can be found at: https://www.TylerDerr.com/

For more information on his current work visit https://tylersnetwork.github.io/

Prof. Callie Hao
Georgia Institute of Technology

Ultra-Low-Latency Graph Neural Networks: Applications and Implementations

In this talk, we present an intriguing application of ultra-low-latency graph neural networks in high-energy physics, necessitating microsecond or even nanosecond-level latency for processing physical events, such as particle collisions. To meet these stringent requirements, specialized hardware and meticulously crafted algorithms are essential. We introduce a series of our hardware architectures, along with design automation, to pursue this goal. Additionally, we highlight several open challenges and opportunities for developing hardware-friendly models for more practical applications.

Biography

Dr. Cong “Callie” Hao is an assistant professor in ECE at Georgia Tech, where she currently holds the Sutterfield Family Early Career Professorship. She received the Ph.D. degree in Electrical Engineering from Waseda University in 2017. Her primary research interests lie in the joint area of efficient hardware design and machine learning algorithms, as well as reconfigurable and high-efficiency computing and agile electronic design automation tools.

For more information on her current work visit https://sites.gatech.edu/ece-callie/

Prof. Sameer Singh
University of California, Irvine

Editing Machines: Updating Text with Up-to-date Information

Reflecting real-world facts and consistent, up-to-date knowledge in text is an important problem. Human-written text may deviate from truth either intentionally, to mislead, or because it becomes obsolete as the world changes. This issue extends to language model-generated text as well, where inconsistencies may arise due to hallucinations or models becoming stale, reflecting an outdated world view. In this talk, I will discuss our work, beginning with the introduction of a novel generation task called faithfully reflecting updated knowledge in text (FRUIT). We will explore the FRUIT-WIKI dataset and EDIT5, a T5-based approach tailored for editing Wikipedia articles with new evidence, highlighting the challenges and opportunities in neural generation models. Next, we will delve into PURR, the Petite Unsupervised Research and Revision model, designed to combat hallucinations in large language models. Through unsupervised training and efficient execution, PURR improves attribution and paves the way for robust editing methods. The talk will conclude with insights on the future of editing models and their transformative potential in managing and trusting information.

Biography

Dr. Sameer Singh is an Associate Professor of Computer Science at the University of California, Irvine (UCI). He is working primarily on the robustness and interpretability of machine learning algorithms and models that reason with text and structure for natural language processing. He has been named the Kavli Fellow by the National Academy of Sciences, received the NSF CAREER award, UCI Distinguished Early Career Faculty award, the Hellman Faculty Fellowship, and was selected as a DARPA Riser. His group has received funding from Allen Institute for AI, Amazon, NSF, DARPA, Adobe Research, Hasso Plattner Institute, NEC, Base 11, and FICO. Sameer has published extensively at machine learning and natural language processing venues and received numerous paper awards, including at KDD 2016, ACL 2018, EMNLP 2019, AKBC 2020, ACL 2020, and NAACL 2022.

For more information on his current work visit https://sameersingh.org/

Prof: Hang Liu
Rutgers University

High-Performance Intelligent Data Analytics

Big data analytics can claim a large share of credit for tackling many grand challenges of our time, such as understanding the spread of pandemics, designing extremely large-scale integrated circuits, and discovering life-saving medicines, among many others. In the High-Performance Data Analytics (HPDA) lab at Rutgers, The State University of New Jersey, we develop High-Performance algorithms and systems to analyze big real-world data, understand the contextual and causal relationships within entities and events, and deliver actionable knowledge to stakeholders in an acceptable time envelope. In this talk, I will share our experiences designing and developing high-performance systems for random walk and graph sampling on large dynamic graphs (featured at Eurosys '23). Subsequently, I will present two follow-up projects: (i) accelerating the training of Graph Neural Networks (GNNs) at scale [SC '23]; (ii) using machine learning to accelerate microarchitecture simulations (featured at Sigmetrics '22 and SC '22); and (iii) offloading the Department of Energy (DoE) flagship direct solvers (i.e., SuperLU) on the FPGA clusters (featured at TPDS).

Biography

Dr. Hang Liu is currently an assistant professor in the Department of Electrical and Computer Engineering at Rutgers, The State University of New Jersey. Prior to this, he was a Presidential Fellowship assistant professor at Stevens Institute of Technology. He received his Ph.D. from George Washington University in 2017 and B.E. from Huazhong University of Science and Technology in 2011. Hang Liu’s research interests include exploiting emerging hardware to build high-performance systems for graph analytics, machine learning, data compression, and numerical simulation. Dr. Liu has received the prestigious NSF CAREER award, IEEE CS TCHPC Early Career Researchers Award for Excellence in High-Performance Computing – 2022, NSF CRII award, the Best Dissertation Award of Electrical and Computer Engineering from the George Washington University, the Champion of GraphChallenge 2018 and 2019, the Lawrence Berkeley National Laboratory SRP fellowship 2019 and 2021, One of the best papers in VLDB '20, and Provost Early Career Award for Research Excellence 2022.

For more information on his current work visit https://asherliu.github.io/