Evaluating derivatives; principles and techniques of algorithmic differentiation, 2d ed.

9780898716597

Evaluating derivatives; principles and techniques of algorithmic differentiation, 2d ed.

Griewank, Andreas and Andrea Walther.

SIAM

2008

438 pages

$73.50

Paperback

QA304

Griewank (mathematics, Humboldt U. of Berlin) and Walther (analysis and optimization of computer models, Dresden Technical U) have thoroughly updated this edition expanding upon the characteristics of algorithmic, or automatic differentiation. They appreciate how this growing area of theoretical research is developing out of concern for accurate and efficient evaluations of derivatives for functional values. They also well understand that the resulting derivative values are useful for all scientific computations that are based on linear, quadratic, or higher orders in nonlinear scalar or vector functions. Therefore they include examples and exercises as they describe tangents and gradients as fundamentals of forward and reverse computation, memory and bounds of complexity, repeating and extending reverse computation, implementation, and software. They also examine Jacobians and Hessians, including observations on efficiency, advances in reversals, including reversal schedules and check pointing, Taylor and tensor coefficients, differentiation without differentiability, and implicit and iterative differentiation.

([c]2009 Book News, Inc., Portland, OR)