No label defined (Q4539016)

For a review of the first ed. (1986) see Zbl 0601.65016 and Zbl 0601.65017. From the preface (translated from the French): We describe the most commonly used methods of matrix computation: (i) the classical methods (Gauss elimination, superrelaxation, QR and Jacobi for eigenvalues, etc.), which are the only ones that appear in every library; (ii) extensions of these methods that constitute improvements (Gauss elimination for sparse matrices, the fast Givens method, etc.); (iii) more recent methods (preconditioned conjugate gradient, Lanczos, multigrid, etc.), developed especially to treat the case of large sparse matrices. Aside from the mathematical exposition of the algorithms and the proof of the convergence theorems, we emphasize the practical evaluation of these algorithms from various aspects: generality of application, precision, stability with respect to round-off errors, computational speed, required memory space, ease of programming, etc. This allows us to compare various methods according to several criteria, whose importance depends on the problem at hand and the type of implementation envisioned. We include the programs of certain algorithms written in (pseudo)scientific language, not in order to replace existing software but rather to show that the programming of the algorithms themselves is often simple. This book is based on a course that we teach at the Conservatoire National des Arts et Métiers, in the computer science section. Its level is that of the second cycle of French higher education. The first sections of several of the chapters (the Gauss method, iterative relaxation methods, iterated powers, etc.) can nevertheless be taught in the first cycle. Other, more recent methods may be used for a third-cycle course. This book is also aimed at engineers and technicians who need to use computers. Here they will find, in accessible language, a synthesis of the numerical methods of linear algebra that are in use today. ''To read this book one needs only an understanding of the basic notions of linear algebra. Moreover, the first chapter contains a review of the principal notions, along with proofs of the least classical results. Numerous exercises, which vary in difficulty and can be solved `with pencil and paper', are included at the end of each chapter, in order to facilitate the assimilation of the material or to give supplementary information''.