Prof. Peter Hoff

Duke University, USA

Title: Adaptive inference for matrices and tensors

Abstract: Multilinear models and low-rank approximations provide the foundation for much of tensor-valued data analysis. But what if the underlying signal is not low-rank or multilinear? In this talk we discuss some techniques for adaptive testing, estimation and confidence interval construction that build upon multilinear models, but do not strictly rely upon multilinear assumptions.

Posted in Speakers.