Sequential Statistical Methods for Reliable Machine Learning @SimonsInstitute
Sequential Statistical Methods for Reliable Machine Learning  @SimonsInstitute
Uploaded October 2026 | Updated November 2026, 3 days ago
Jeremias Sulam (Johns Hopkins University)
https://simons.berkeley.edu/talks/jeremias-sulam-johns-hopkins-university-2026-10-06
Trustworthy AI: From Hallucinations to Reliable Autonomy

Modern machine learning models are increasingly deployed in dynamic environments where reliability must be established continuously rather than assumed. In this talk, I will present two recent advances in sequential statistical inference that address complementary aspects of this challenge. I will first introduce efficient sequential hypothesis tests for continuously auditing deployed machine learning systems, enabling rapid detection of performance degradation while maintaining rigorous Type I error guarantees. I will then show how similar sequential perspectives lead to parameter-free online conformal prediction methods that provide adaptive, group-conditional uncertainty quantification under evolving data distributions. Together, these results illustrate how sequential inference provides a principled statistical framework for moving from detecting when models fail to quantifying when individual predictions can be trusted.
Sequential Statistical Methods for Reliable Machine LearningNodal surplus and the spectral decomposition of the cycle spaceSturmian discrete and metric graphs - appearance and absence of gapsPseudorandomness against AM circuits and new oracle separationsWhat Really Separates Autoregression and Diffusion? A Synthesis and Path BeyondWhen Can We Trust What a Model Has Learned? Identifiability, Causality, and Reliable AIGround Energy estimation of Quantum Impurity model is in BQPStructure vs. Randomness (Part 2)Graphs versus Manifolds in Eigenvalues and Distances Between SubsetsDerandomizing Space-Bounded Computation (Part 1)Probability, Information, and DiscrepancyCondensers/extractors for block sources survey
Simons Institute for the Theory of Computing |

Sequential Statistical Methods for Reliable Machine Learning

SHARE TO X SHARE TO REDDIT SHARE TO FACEBOOK WALLPAPER