Anatomy of Optimizing an Algorithm: A Look at the SVD Algorithm

Jack Dongarra (University of Tennessee)

Frank Adams 1, Alan Turing Building,

In this talk we will examine the progress made in improving the performance
of the singular value decomposition algorithm from sequential
implementations, to blocked reduction, to parallel implementations, to
accelerator based usage.  The singular value decomposition provides a
convenient way for breaking down a matrix into simpler, meaningful
components. The singular value decomposition of a matrix has many practical
uses such as principal component analysis, data compression, and
information retrieval.
Import this event to your Outlook calendar
▲ Up to the top