This project establishes a new institute on the Foundations of Data Science at the University of Texas at Austin. The Institute will be a collaboration between eight PIs in the electrical engineering, computer science, mathematics and statistics departments at UT Austin, as well as postdocs and graduate students from the new programs this Institute establishes. It will form a central hub for theoretical research into machine learning and data science by looking at foundational approaches to analysis and design. This is necessary to devise novel complex and sophisticated machine-learning and artificial-intelligence theory and algorithms that can handle the accelerating scale of received data and the faster computational speeds of computers. The algorithms and systems will interpret and predict behavior from data and the environment with the goal towards better design methods performed in a principled way. The research will also open avenues for applications in fields such as autonomous vehicles and personalized medicine. The research and education will be integrated to create new inter-departmental postdoctoral and graduate research programs, establish a unified degree and portfolio program in data science at UT Austin, run dedicated seminar series and hold workshops, and partner with industry as well as domain experts in the sciences. It will significantly expand, via funded initiatives, the PIs' ongoing efforts to expand participation of under-represented groups in this important field.
Research focuses on fundamental mathematical theory of machine learning and optimization, including neural networks, robustness, and graphs. The research is organized around three themes: (a) developing an algorithmic theory for deep learning, with new and provable methods for training, doing hyper parameter optimization and developing confidence measures, (b) making machine learning robust to both adversarial and incidental errors in data, and (c) devising new methods for statistical inference using graph algorithms, including fast estimation of graph statistics, and their use in biological and vision applications.
This project is part of the National Science Foundation's Harnessing the Data Revolution (HDR) Big Idea activity.
This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.
