NSF-BSF: CIF: Small: Statistical Learning and Inference via Innovations

This NSF-BSF project develops theoretical and algorithmic foundations for online learning and decision-making involving sequential data under unknown stochastic models. Along with theoretical investigation and algorithmic developments, this project has a significant application component on real-time monitoring and control of critical infrastructure networks. Specific applications include probabilistic forecasting of renewable energy and electricity prices, and detecting emerging behaviors such as those induced by faults or cyber-attacks in information infrastructure. This research contributes to foundational technologies critical to the nation’s power and information infrastructures and fosters international collaborations and industry partnerships. Undergraduate and graduate education activities broaden the impact of this project. The research focuses on developing holistic approaches to integrating representation learning with real-time inference and decision-making. The technical approaches are rooted in classical foundations of statistical inference, advancing some of the most powerful model-based ideas of innovation representation, sequential decision-making, and distributed filtering with modern data-driven generative AI solutions to overcome critical barriers arising from unknown, nonparametric, high-dimensional time series models. Research activities are structured under three thrusts: (i) developing a theoretical and algorithmic foundation for representation learning of nonlinear and nonparametric time series models, (ii) developing statistical inference and learning methodologies in both centralized and distributed settings for sequential data under unknown nonparametric stochastic models, and (iii) applying, validating, and evaluating developed solutions to critical infrastructure monitoring using field collected real-time data. Tightly integrated with the research activities are extensive collaborations with the leading power industry and cybersecurity industry, in particular, the IBM Cybersecurity Center through the NSF-BSF collaboration. This project also enriches undergraduate and graduate curriculum by developing experimental courses to provide hands-on research experiences to students. 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.