Urban AI Seminar Series

Title:
Graph Representation Learning for Network Generation, Optimization, and Verbalization

Seminar Day:
May 23, 2025

Presenter:
Liang Zhao
Associate Professor
Department of Computer Science
Emory University

Abstract:
Graphs are ubiquitous data structures that denote entities and their relations, such as social networks, citation graphs, and neural networks.  The topology of graphs is discrete data, which prevents it from enjoying numerous mathematical and statistical tools that require structured data.  Graph representation learning aims to map graphs to their vector representations without substantial information loss, hence paving a new pathway for solving graph problems without discrete algorithms.  In this talk, we will first introduce our recent works on graph representation learning that can preserve graphs’ geometric information and properties.  Then, we will exemplify several interesting research areas where their problem-solving benefits from our leveraging of graph representations.  The first area is to solve graph optimization problems, such as influence maximization, source localization, etc., using continuous optimization over graph representations.  The second area is to capture and predict deep learning models’ dynamics over data distribution drifts, where the graph representation of neural networks is learned to reflect their functional space.  The third area is to investigate the correlation and difference of the two views of graphs in mathematical language and natural language, where the graph representation acts as their bridge, with the help of large language models.

Speaker:
Dr. Liang Zhao is an associate professor in 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 PhD in 2016 from the Computer Science Department at the Virginia Polytechnic Institute and State University 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, language and multimodal foundation models, distributed optimization, and interpretable machine learning.  He won the National Science Foundation Career Award and the Middle-Career Award from the Institute of Electrical and Electronics Engineers Computer Society on Smart Computing.  He also obtained many prestigious awards from industry, such as the Meta Research Award, Amazon Research Award, Cisco Faculty Research Award, and Jeffress Trust Award.  He was recognized as one of the “Top 20 Rising Star in Data Mining”, by Microsoft Search in 2016.  He has won several best paper awards and shortlists.  He was recognized as a “Computing Innovative Fellow Mentor”, in 2021 by the Computing Research Association.

Title:
EPRI READi Inititaive

Seminar Day:
July 16, 2024

Presenter:
Stephen Signore
Performance Monitoring and Diagnostic
Tennessee Valley Authority

Abstract:
Dr. Signore currently processes for Tennessee Valley Authority’s (TVA) River Forecast Center will be discussed including reactions to severe storm events and seasonal operations that provide multipurpose benefits to the region.  TVA’s Probabilistic Flood Risk Analysis tool will be discussed and how that can be used to inform decision makers of the potential outcomes of severe storm events.  The real time coordination between the RFC and the Performance Monitoring & Diagnostics (PM&D) group that Dr. Signore is a part of and the other responsibilities of the PM&D group in safe, efficient, reliable operations of TVA assets will be covered.  TVA’s internal work on reliability analyses of hydroelectric powertrain assets that resulted in increased funding for major maintenance work and led to Dr. Signore’s ongoing coordination with an Oak Ridge National Laboratory (ORNL) research group’s efforts to advance the area will be presented as well.

Speaker:
Stephen Signore earned his Ph.D. in Energy Science & Engineering from UT-Knoxville in 2017 and holds B.S. Civil Engineering and Environmental Engineering degrees from Florida State University.  After completing a post-doctoral appointment at ORNL in the Environmental Sciences Division, Stephen joined the Performance Monitoring & Diagnostics team at Tennessee Valley Authority in April 2020.   His current work focuses on identifying long term trends in hydropower equipment monitoring data that can lead to actionable decisions to reduce wear and tear on the units to improve system reliability and implementing changes to hydro unit characteristics that interact with how TVA’s hydropower fleet is scheduled.  He also serves as a point of contact for TVA for our partnership with ORNL’s Hydropower Fleet Intelligence project and is active in the Electric Power Research Institute Renewable Energy Accelerated Deployment Initiative (READi).

Title:
Quantifying Role of Groundwater in Global Water Resources within Context of Extremes

Seminar Day:
June 6, 2024

Presenter:
Bridget R. Scanlon
Bureau of Economic Geology
Jackson School of Geosciences
University of Texas at Austin

Abstract:
Groundwater is playing an increasingly important role in global water resources with intensification of extremes. Quantifying the importance of groundwater and its role as a buffer is critical for developing sustainable water management practices. Our analysis of global water trends benefits from advances in remote sensing, especially GRACE satellite data, global and regional modeling, and expanding monitoring networks.

GRACE satellite data show declining, stable and rising trends in total water storage over the past two decades in various regions globally. The visual power of GRACE data helped communicate water storage variability to the public and influence water policy in many regions, including India and the US.

Groundwater monitoring provides longer-term context over the past century, showing rising water storage in northwest India, central Pakistan and the northwest United States, and declining water storage in the US High Plains and Central Valley. variability causes some changes in water storage, but human intervention, particularly irrigation, is a major driver. Water-resource resilience can be increased by diversifying management strategies. These approaches include green solutions, such as forest and wetland preservation, and grey solutions, such as increasing supplies (desalination, wastewater reuse), enhancing storage in surface reservoirs and depleted aquifers, and transporting water. A diverse portfolio of these solutions, in tandem with managing groundwater and surface water as a single resource, can address human and ecosystem needs while building a resilient water system.

Speaker:
Bridget Scanlon is a Research Professor at the Bureau of Economic Geology, Jackson School of Geosciences, University of Texas at Austin. Her current research focuses on various aspects of water resources, including global assessments using satellites and modeling, management related to extremes, and water energy interdependence. She has authored ~ or co-authored ~170 publications. Dr. Scanlon is an AAAS, AGU and GSA Fellow and NAE member.

Title:
Defining and Practicing Computational Urban Science

Seminar Day:
February 1, 2024

Presenter:
Xinyue Ye
Professor, Urban Informatics
Texas A&M University

Abstract:
Big data and computational algorithms have become increasingly integrated into the built environment and our daily lives, leading to a marked rise in computational urban science. Understanding the multi-dimensional interplay between computation and the spatial-social aspects of urban life carries both theoretical and practical implications for urban planning and management. Numerous analytical methods have been developed and applied to handle high-dimensional, heterogeneous, and unstructured location-based social data from urban settings. Computational urban science comprises four interdependent layers: human dynamics-centered, platform-based, action-oriented, and convergence-driven. As a research paradigm rooted in computational thinking and spatiotemporal synthesis, computational urban science offers a vital framework for addressing many of the urgent challenges in urban sustainability from a systematic perspective. This framework can foster a collective understanding of the current state of urban infrastructure and showcase innovative approaches to enhance urban resilience and efficacy beyond mere technological integration. This presentation will also highlight relevant research projects and activities undertaken by the Center for Geospatial Sciences, Applications and Technology, the Texas A&M Institute of Data Science-Urban AI Lab, and the Journal of Computational Urban Science.

Speaker:
Xinyue Ye is the Harold Adams Endowed Professor in Urban Informatics and Stellar Faculty Provost Target Hire at Texas A&M University (TAMU). He serves as the Faculty Fellow of Strategic Initiatives and Partnerships for The Division of Research at TAMU. With a career experience in urban planning, regional economics, geographic information systems, and computational science, his research focuses on geospatial artificial intelligence, smart cities, and convergence research. Dr. Ye models the space-time perspective of socioeconomic inequality and human dynamics for applications in various domains. Due to his innovative research integrating geography, planning, and computational science, Dr. Ye was the most junior faculty member and only the second planning faculty elected as a Fellow of the American Association of Geographers in 2022. His work has been funded by the National Science Foundation, National Institutes of Health, National Institute of Justice, National Academies, National Aeronautics and Space Administration, National Oceanic and Atmospheric Administration, Department of Commerce, Department of Energy, Department of Housing and Urban Development, Department of Transportation, Department of Health and Human Services, Microsoft, and the Canada Social Sciences and Humanities Research Council.

Title:
Collaborative Virtual Augmented Reality

Seminar Day:
November 10, 2023

Presenter:
Bin Li
Associate Professor
Department of Electrical Engineering
Pennsylvania State University

Abstract:
Virtual reality (VR) over wireless networks can provide an interactive and immersive experience for multiple users simultaneously and thus has many applications, especially in VR-based education/training. However, satisfactory personalized user experience in such wireless immersive services demands stringent performance requirements, including: (1) high-speed and high-resolution panoramic image rendering; (2) extremely low delay guarantees; and (3) seamless user experience. Besides the aforementioned requirements, collaborative user experience requires scalability of VR service. In this talk, we will talk about our network algorithm design for providing both personalized and scalable collaborative VR experiences over wireless networks.

Speaker:
Bin Li is currently an associate professor in the Department of Electrical Engineering at Pennsylvania State University. He received Ph.D. degree in Electrical and Computer Engineering from The Ohio State University (OSU). He was a Postdoctoral Researcher in the Coordinated Science Lab (CSL) at the University of Illinois at Urbana-Champaign (UIUC). His research focuses on the intersection of networking, machine learning, and system development, and their applications in networking for virtual/augmented reality, mobile edge computing, mobile crowdsourcing, and Internet-of-Things (IoT). He received both the National Science Foundation (NSF) CAREER Award and Google Faculty Research Award.

Title:
The Next Big Leap: How AI and Data Science Can Transform Disaster Resilience Research and Practice

Seminar Day:
July 28, 2023

Presenter:
Ali Mostafavi,
Zachry Endowed Associate Professor
Department of Civil and Environmental Engineering
Texas A&M University

Abstract:
This talk focuses on smart resilience, which is a new paradigm for harnessing community-scale big data and Artificial Intelligence (AI) to augment disaster resilience capabilities such as risk prediction, rapid impact assessment, predictive infrastructure failure monitoring and situational analysis.  The big data revolution is transforming the science and practice of resilience to crises.  Resilience in this context is defined as the ability of various systems (e.g., infrastructure, businesses, emergency response, and critical facilities) to maintain functionality needed for residents to reduce various social, economic, physical, and well-being impacts of floods.  Data science and AI solutions could significantly augment the capabilities needed for smart flood resilience.  In this talk, an overview of a number of studies conducted at the UrbanResilience.AI Lab will be presented in which various modes of big data such as location-based, digital trace, social media, crowdsourced, physical sensor data, and satellite images have been collected and analyzed in creating machine learning and deep learning models to improve different disaster resilience capabilities.  The talk also highlights opportunities to promote equity in disaster resilience processes using responsible AI and data science opportunities. Also, challenges of responsible data science in this field will be discussed.