scientific article; zbMATH DE number 1383467

From MaRDI portal
Publication:4719212





The normalized QR algorithm forms the core of many computations in signal processing and system identification. The authors provide a Bayesian analysis of a parallel version of the algorithm. The Bayesian approach allows them to get insight into the numerical properties of the algorithm in terms of probability distributions of the data propagated in the normalized QR filter. From these, the authors derive probability distributions of the number of unused bits in fixed point arithmetic, which provides a basis for optimal selection of word-length for fixed-point arithmetic. The paper reviews in a fairly self-contained manner the ingredients of the proposed methodology and may serve as a starting point for research in other settings.











This page was built for publication:

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q4719212)