Evaluating Derivatives. Principles and Techniques of Algorithmic Differentiation
Buy a book... In Association with Amazon.co.uk
Author(s): A. Griewank and A. Walther
Publisher: SIAM
ISBN: 9780898716597
Format: softback
438pp
Price: $73.50
Review Date: 21 November 2008
Review: This book is intended for designers of algorithms and software for nonlinear computational problems as well as users of numerical software who need to provide derivative values, sparsity patterns and other dependence information. The reader should find the principles and techniques here very useful for any kind of nonlinear modelling in scientific computing. The book is in three parts: I. Tangents and gradients: framework for evaluating functions, fundamentals of forward and reverse, memory issues and complexity bounds, repeating and extending reverse, implementation and software. II Jacobians and Hessians: sparse forward and reverse, exploiting sparsity by compression, going beyond forward and reverse, Jacobian and Hessian accumulation, observation and efficiency. III Advances and Reversals: Taylor and tensor coefficients, differentiation without differentiability, implicit and iterative differentiation.