Interior point gradient algorithm for totally nonnegative least-squares problems in inequality sense

Abstract

In this paper we study a scaled interior-point gradient algorithm for totally nonnegative least-squares problems in the inequality sense. The global convergence of the algorithm is established. Finally, extensive computational comparison with the so-called generalized Newton algorithm is given showing the superior performance of the gradient-based algorithm.

Keywords:

• linear inequalities
• least-squares problems
• gradient methods
• interior-point methods
• generalized Newton method

2010 AMS Subject Classifications::

• 90C25
• 90C51
• 49M15