Tensor decompositions play a key role in learning the parameters of latent variable models. In this tutorial, I will introduce Jennrich's algorithm, which is a simple and somewhat forgotten algorithm. Indeed it has been rediscovered at least half a dozen times, by my count. With this as a starting point, I will give a unified exposition of some of the algorithmic applications of tensor decompositions, ranging from phylogenetic reconstruction to learning mixture models to orbit retrieval.
Copyright. All Rights Reserved.