UCPH Statistics Seminar: Alexander Taveira Blomenhofer
Speaker: Alexander Taveira Blomenhofer from QMATH
Title: Algebraic identifiability of mixture models
Abstract: Gaussian mixture models are ubiquitous in statistical learning, e.g., as a standard model for clustering. Central questions on mixture models are about parameter identifiability and estimation algorithms. Recent results on these questions have used techniques from tensor geometry. This talk gives an overview of results and techniques at the interface of mixture models and tensor geometry.