The 1996 conference on the State of the Art in Numerical Analysis was organized to provide the numerical analysis community, and users of numerical methods, with a forum where an account of the important recent developments in the subject could be presented in a coherent and concentrated way in a manner accessible to the non-specialist in the sub-area. This volume extends to a much wider audience the opportunity given to those who attended the meeting. It contains full versions of all the papers presented, with one exception on Neural Nets. In addition to recent developments in mainstream topics, linear algebra, ordinary differential equations, approximation and optimization, it contains papers devoted to two important application areas, numerical tomography and image processing.Compare with the corresponding 1986 volume, some topics show a continuous and natural development, while others show significant departure from conventional trends. A recurring theme is the solution of large problems and exploitation of structure. Underlying many of the developments is the fact that increasingly complicated and sophisticated problems are now amenable to an increasingly powerful range of numerical techniques. This is greatly helped by the advent of excellent computer languages, like MATLAB, and state of the art Fortran software such as LAPACK, so that the fundamental building blocks for much of the armoury of a numerical analysis are now readily accessible.