RALF GIERING
MS 300-323
Jet Propulsion Laboratory
4800 Oak Grove Drive
Pasadena CA 91109, USA
http://puddle.mit.edu/~ralf
Ralf.Giering@jpl.nasa.gov
Abstract:
Adjoint models are powerful tools,
they are increasingly being used for sensitivity studies,
data assimilation, parameter estimation, and more.
The adjoint model efficiently computes the sensitivity of
an output variables with respect to arbitrarily many input variables.
The construction of adjoint models by hand is tedious and error prone. Automatic differentiation reduces this work substantially. The basics of Automatic Differentiation, its advantages, and its limitations will be presented.
The Tangent linear and Adjoint Model Compiler (TAMC) is a source-to-source tool that generates tangent linear and adjoint models from Fortran code. The performance of TAMC generated code will be compared to hand coded counterparts for functions in the MINPACK-2 collection. Adjoint models have been constructed by TAMC for large scale applications in dynamic meteorology and oceanography; an implemented checkpointing technique reduces the memory requirements significantly. Implementation and performance of some adjoint models will be discussed.