Automatic Differentiation with TAPENADE

Laurent Hascoet Valerie Pascual Alain Dervieux INRIA, BP93, 06902 Sophia-Antipolis, France

Book chapter, conclusions of the "Aeroshape" project (10 pages)

Abstract: We present TAPENADE, a tool for Automatic Differentiation (AD). AD transforms a program that computes or simulates a mathematical vector function into a new program that computes derivatives of this function. Gradients and Adjoints are among the most popular derivatives, as they are required in optimization. We therefore focus on the reverse mode of TAPENADE, which returns gradients. Our goal here is to present TAPENADE so that end-users can understand the differentiated programs better, and therefore can use this tool with increased confidence. In particular, we describe the underlying model of the reverse mode of TAPENADE, and we present quickly its user interface with links to tutorial and reference manual.

Keywords: Automatic Differentiation, Algorithmic Differentiation, Adjoint, Gradient, Optimization, Inverse Problems, Static Analysis, Data-Flow Analysis, Compilation

Full text (pdf)


@inBook{HPD05,
  author = {Hasco\"et, L. and Pascual, V. and Dervieux, D.},
  title = {Automatic Differentiation with TAPENADE},
  booktitle = {Notes on Numerical Fluid Dynamics},
  editor = {Selmin, V.},
  publisher = "Springer",
  note = {to appear},
  year=2005
}