Introduction to Derivative-Free Optimization (Mps-Siam by Andrew R. Conn, Katya Scheinberg, Luís N. Vicente
By Andrew R. Conn, Katya Scheinberg, Luís N. Vicente
The absence of derivatives, frequently mixed with the presence of noise or loss of smoothness, is a big problem for optimization. This publication explains how sampling and version innovations are utilized in derivative-free tools and the way those equipment are designed to successfully and carefully resolve optimization difficulties. even if with no trouble available to readers with a modest historical past in computational arithmetic, it's also meant to be of curiosity to researchers within the box. advent to Derivative-Free Optimization is the 1st modern finished remedy of optimization with no derivatives. This publication covers many of the correct periods of algorithms from direct seek to model-based methods. It includes a entire description of the sampling and modeling instruments wanted for derivative-free optimization; those instruments let the reader to higher comprehend the convergent houses of the algorithms and establish their variations and similarities. creation to Derivative-Free Optimization additionally comprises research of convergence for transformed Nelder Mead and implicit-filtering equipment, in addition to for model-based tools comparable to wedge tools and strategies in response to minimum-norm Frobenius versions. viewers: The ebook is meant for an individual drawn to utilizing optimization on difficulties the place derivatives are tricky or very unlikely to acquire. Such audiences comprise chemical, mechanical, aeronautical, and electric engineers, in addition to economists, statisticians, operations researchers, administration scientists, organic and scientific researchers, and computing device scientists. it's also acceptable to be used in a complicated undergraduate or early graduate-level path on optimization for college students having a history in calculus, linear algebra, and numerical research. Contents: Preface; bankruptcy 1: advent; half I: Sampling and modeling; bankruptcy 2: Sampling and linear versions; bankruptcy three: Interpolating nonlinear versions; bankruptcy four: Regression nonlinear types; bankruptcy five: Underdetermined interpolating versions; bankruptcy 6: making sure good poisedness and appropriate derivative-free versions; half II: Frameworks and algorithms; bankruptcy 7: Directional direct-search tools; bankruptcy eight: Simplicial direct-search tools; bankruptcy nine: Line-search equipment in keeping with simplex derivatives; bankruptcy 10: Trust-region equipment according to derivative-free versions; bankruptcy eleven: Trust-region interpolation-based equipment; half III: evaluation of alternative themes; bankruptcy 12: evaluation of surrogate version administration; bankruptcy thirteen: assessment of limited and different extensions to derivative-free optimization; Appendix: software program for derivative-free optimization; Bibliography; Index.