UD Engineering’s Guangmo (Amo) Tong receives NSF CAREER award for computer science research
From self-driving cars to facial recognition software, technologies powered by artificial intelligence rely on powerful algorithms that can make sense of complex, real-world data. These algorithms are part of a larger computational workflow, known as a pipeline, that involves collecting data, identifying patterns and making decisions on what to do next.
But decision-making pipelines that are specialized for one type of task — to recognize faces, for example — can’t be used to analyze other types of information, such as helping a smart car decide whether to turn right or left at the next intersection. Given these limitations, how can researchers create new algorithms and decision-making pipelines that can work across a wider variety of datasets and real-world scenarios?
Guangmo (Amo) Tong, assistant professor in the Department of Computer and Information Sciences at the University of Delaware’s College of Engineering, is using fundamental computer science research toward improving data-driven decision making. Through a better understanding of the theories that underlie algorithmic decision-making, Tong’s research can support future breakthroughs across a wide range of applications.
Research in Tong’s Computational Data Science Lab focuses on two areas of computer science: Designing algorithms, the step-by-step instructions used by computer programs to complete specific tasks, and developing statistical learning techniques, the types of mathematical models that are used to analyze and interpret large datasets.
This image shows an example of the decision-making pipelines that will be developed as part of Tong’s CAREER award.
Beyond focusing on the application of algorithms and statistical models, Tong’s group is focused on gaining a more thorough understanding of how algorithms make decisions using data. “We need a deeper understanding of decision-making, and this foundation can in turn improve state-of-the-art solutions for specific applications,” said Tong. “We believe that the foundation is important in that it justifies the model performance by providing provable guarantees.”
Tong’s expertise in computational data science was recently recognized by the National Science Foundation (NSF) with a Faculty Early Career Development Program (CAREER) award, a prestigious grant that supports early-career faculty who have the potential to become leaders in both research and education. Tong’s CAREER award includes $513,468 in funding for five years and began in September of 2022.
“Different from many researchers in the community that focus on how to apply existing AI techniques to different scenarios, Dr. Tong aims to tackle a fundamental question on how to design and implement general decision-making algorithms that can potentially work across a variety of datasets,” said Weisong Shi, professor, CIS department chair and a collaborator with Tong on this award. “This task is extremely challenging but will have a much broader impact.”
Tong’s project will focus on developing decision-making pipelines for discrete decisions, or ones that have a specific action associated with them (such as “turn right” or “stop”). This research will focus on both algorithm design as well as exploring which statistical models are the most appropriate given the type of data that is being analyzed.
“In the beginning, we plan on focusing more on the theoretical parts of this research — designing algorithms to solve optimization problems and learning pipelines,” said Tong. “After that, we plan to focus on several applications, including social network analysis and autonomous systems.”
In the area of social network analysis, Tong and his group will use this award to continue their work on “social contagion management,” or how different messages move through a network. By figuring out how information flows through social networks, Tong’s research could help researchers understand the factors that influence the dynamics of messages, behaviors, and opinions.
Tong’s group will also use the award to help Gonzalo Arce, the Charles Black Evans Professor in the Department of Electrical and Computer Engineering, address challenges around inverse problems in complex networks, computational imaging, and 3D Lidar sensing in his lab.
This is an example of a social network (left) and the diffusion of an information cascade (right). Tong and his team will be studying both of these structures as part of their work on “social contagion management,” or how different messages move through a network.
“One of the things we are working on is developing hypergraph neural networks, and what is important is to have a theoretical model of how powerful and expressive these networks are,” said Arce. “Our group is developing network architectures and algorithms that drive these, but it’s nice to have his group analyze how much more powerful these new networks are compared to current state-of-the-art.”
Researchers in Tong’s Computational Data Science Lab will also continue testing their newly developed algorithms on autonomous driving platforms such as HydraOne, an autonomous vehicle platform designed by Shi’s lab, to see how well their algorithms work in the real world. Thanks to this award, Tong is planning to get both undergraduate and graduate students involved in this and other related projects that take his group’s research findings from theoretical concepts to applied settings.
The award will also support Tong’s ongoing outreach activities, with continued high school student mentorship in collaboration with UD’s K-12 engineering outreach program, and will also support curriculum development for computer science graduate education.
Tong said that instructing graduate students in the fundamentals of algorithm design and statistical modeling is key for training the next generation of data scientists and software engineers, who will work with increasingly large datasets that require new algorithms, statistical learning models and computational approaches to make sense of complex, real-world data.
“It is often possible to implement algorithms in applications without understanding why this works, but I hope that I can leverage these projects to help students focus more on the foundations of machine learning and algorithms with a focus on curriculum design,” said Tong.
Arce added that a strong foundation in theory also opens new areas of collaboration at UD. “In machine learning, having a well-rounded portfolio is important, and having the applications, the theory and the underlying architectures is the ideal package,” he said.
“Dr. Tong’s recognition is the second NSF CAREER award the CIS department has received this year, after Dr. Lena Mashayekhy received her award in April,” said Shi. “Together, nearly 40% of our tenure-track and tenured faculty members in the department have received this prestigious award, demonstrating the high quality of cutting-edge research by our faculty and students at the University of Delaware.”
Article by Erica K. Brockmeier | Photo by Kathy F. Atkinson | Illustrations by Joy Smoker | November 17, 2022