Adaptive Robust Confidence Intervals in High-Dimensional Statistics

Watch a 47-minute mathematical statistics lecture exploring the construction of confidence intervals under Huber's contamination model, presented at the Centre International de Rencontres Mathématiques. Dive deep into the analysis of adaptation costs when contamination proportion is unknown, with specific focus on Gaussian location models where adaptive confidence intervals are proven to be exponentially wider than non-adaptive ones. Explore the implications for general location models and examine the same problem in network settings, particularly for Erdos-Renyi graphs with node contamination. Learn how the complexity of adaptive confidence interval construction relates to detection thresholds between Erdos-Renyi model and stochastic block model. Access this recording from the "Meeting in Mathematical Statistics: New challenges in high-dimensional statistics" conference, complete with chapter markers, keywords, abstracts, and bibliographies through CIRM's Audiovisual Mathematics Library.