ABS algorithms for linear equations and optimization (Q1593821): Difference between revisions

The authors review the basic properties of Abaffy-Broyden-Spedicato (ABS) methods for linear systems and some of their applications to optimization problems. Many subclasses of the ABS class are described : the conjugate directions subclasses, the orthogonally scaled subclass, the Huan algorithm, the implicit LU algorithm, etc., by choosing the parameters of the ABS methods. The quasi-Newton methods, under some conditions, can be embedded in the ABS class. As applications to optimization, the unconstrained optimization case is discussed. The solutions of Kuhn-Tucker equations can be written in ABS form. Also, the simplex method for linear programming in standard form is reformulated via the implicit LX algorithm. The paper has a descriptive form, without theorems, lemmas, definitions.