The Newton-Cauchy framework. A unified approach to unconstrained nonlinear minimization (Q1317214)

This book gives a unified presentation of the methods of Cauchy and Newton, their generalizations and implementations. It consists of six chapters. First the conjugate gradient and quasi-Newton methods are introduced in the framework of convex quadratic minimization. The idea here is to motivate the methods and demonstrate their properties in a particular simple setting. The ``metric-based'' method of Cauchy and the ``model-based'' method of Newton are then studied in the next two chapters for general nonlinear functions. Chapter 4 ties together the preceding topics, develops the Newton/Cauchy framework and indicates its rich array of implications. The last two chapters overview nonlinear unconstrained methods and discuss the basic conditions for establishing global convergence of algorithms derived from this framework.