Mathematics Colloquia and Seminars

Return to Colloquia & Seminar listing

Phd Exit Seminar: Overcomplete tensor decomposition: algorithm and application

Special Events

Speaker: Haolin Chen
Location: Zoom
Start time: Tue, Jul 19 2022, 2:00PM

Tensor decomposition has been a useful tool in data analysis for years. Recently it finds applications to estimating the parameters of various latent variable models. In this talk, we will discuss overcomplete tensor decomposition, i.e., the components of the tensor are not linearly independent. We will present an efficient, randomized algorithm for overcomplete tensor decomposition, with applications to a family of blind deconvolution problems of probability distributions, and specifically, the learning problem of a family of Gaussian mixture models.