Title: Solving least squares problems by preconditioned iterative methods.
Abstract:
The efficient solution of large linear least squares problems in which
the system matrix contains rows with very different densities is
challenging. There have been many classical contributions to solving this
problem that focus on
direct methods; they can be found in the monograph by Ake Bjorck.
Such solvers typically perform a splitting of the input matrix rows.
In this talk, we discuss a number of approaches as, for example,
different processing of rows in the preconditioner construction,
Schur complement reduction as well as matrix stretching
in which dense rows are replaced by submatrices with much sparser rows.
Experimental problems demonstrate not only strengths but also limitations of
some of these approaches.
