56:50
published on April 10, 2025
Deep Out-of-the-distribution Uncertainty Quantification in for Data (Science) Scientists
By Nicolas Vayatis
Bayesian multiple testing for dependent data and hidden Markov models - lecture 1
Appears in collection : End-to-end Bayesian Learning Methods / Solutions de bout-en-bout en apprentissage Bayésien
Hidden markov models (HMMs) have the interesting property that they can be used to model mixtures of populations for dependent data without prior parametric assumptions on the populations. HMMs can be used to build flexible priors. I will present recent results on empirical Bayes multiple testing, non parametric inference of HMMs and fundamental limits in the learning of HMMs.
