Quadratic Envelope Regularization for Structured Low Rank Approximation

Abstract

Compressed sensing techniques, such as nuclear norm minimization, can be used for structured low rank approximation, but it is well known that these methods lead to suboptimal results. In this article we consider how to improve this approach by use of so called "quadratic envelopes". The new feature is the extension to weighted matrix spaces relying on tensors, and we show how this can be used for improved accuracy of complex frequency estimation methods.