M. Voßbeck, R. Giering, and T. Kaminski. Development and First Applications of TAC++. In C. Bischof, H. M. Bücker, P. D. Hovland, U. Naumann, and J. Utke, editors, Advances in Automatic Differentiation, Lecture Notes in Computational Science and Engineering, pages 187-197, Berlin, 2008. Springer. [ .pdf ]
T. Pock, M. Pock, and H. Bischof. Algorithmic differentiation: Application to variational problems in computer vision. IEEE Transactions on Pattern Analysis and Machine Intelligence, 29(7):1180-1193, 2007. [ .pdf ]
R. Giering and T. Kaminski. Automatic sparsity detetection implemented as source-to-source transformation. In Vassil N. Alexandrov, Geert Dick van Albada, Peter M. A. Sloot, and Jack Dongarra, editors, Computational Science - ICCS 2006, volume 3394 of Lecture Notes in Computer Science, pages 591-598, Heidelberg, 2006. Springer. [ DOI | .pdf ]
R. Giering, T. Kaminski, and T. Slawig. Generating Efficient Derivative Code with TAF: Adjoint and Tangent Linear Euler Flow Around an Airfoil. Future Generation Computer Systems, 21(8):1345-1355, 2005. [ DOI | http | .pdf ]
We give a TAF overview and describe in some detail, how iterative solvers can be handled efficiently. As an application we demonstrate the generation of adjoint, tangent linear, and Hessian code for a CFD solver in an airfoil configuration. We also list a few other large-scale applications and their performance.
T. Kaminski, R. Giering, and M. Voßbeck. Efficient sensitivities for the spin-up phase. In H. M. Bücker, G. Corliss, P. Hovland, U. Naumann, and B. Norris, editors, Automatic Differentiation: Applications, Theory, and Implementations, volume 50 of Lecture Notes in Computational Science and Engineering, pages 283-291. Springer, New York, NY, 2005. [ DOI | .pdf ]
Demonstrates an alternative AD strategy for iterative solvers that evaluates the full Jacobian for the final iteration.
M. Voßbeck, R. Giering, and T. Kaminski. Towards a tool for forward and reverse mode source to source transformation in C++. In M. Bücker, G. Corliss, P. Hovland, U. Naumann, and B. Norris, editors, NOT accepted for AD 2004, Lecture Notes in Computational Science and Engineering, Berlin, 2004. Springer. [ .pdf ]
A feasibility study for our C++ AD-tool. A short 129 C-line Roe solver is differentiated in reverse mode. The automatically generated adjoint code runs about three times slower than the solver itself.
E. Slusanschi and H. M. Bücker. On the Limits of Current Implementations of Algorithmic Differentiation. In D. Petcu, V. Negru, D. Zaharie, and T. Jebelean, editors, Proceedings of the 6th International Symposium on Symbolic and Numeric Algorithms for Scientific Computing, SYNASC04, Timisoara, Romania, September 26-30, 2004, pages 295-306, Timisoara, 2004. MIRTON. [ .html ]
R. Giering and T. Kaminski. Applying TAF to generate efficient derivative code of Fortran 77-95 programs. PAMM, 2(1):54-57, 2003. [ http | .pdf ]
Here we give an overview on how TAF approaches typical challenges of AD such as handling of badly written program code, of large memory/disk requirements, of iterative solvers or of black box routines. We also point out, where the user is required to prepare his program code prior to invoking TAF.
T. Kaminski, R. Giering, M. Scholze, P. Rayner, and W. Knorr. An example of an automatic differentiation-based modelling system. In V. Kumar, L. Gavrilova, C. J. K. Tan, and P. L'Ecuyer, editors, Computational Science - ICCSA 2003, International Conference Montreal, Canada, May 2003, Proceedings, Part II, volume 2668 of Lecture Notes in Computer Science, pages 95-104, Berlin, 2003. Springer. [ .ppt/.pdf | .pdf ]
The paper presents a prototype of a Carbon Cycle Data Assimilation System (CCDAS), which is composed of a terrestrial biosphere model (BETHY) coupled to an atmospheric transport model (TM2), corresponding derivative codes as well as a derivative-based optimisation routine. In calibration mode, we use first and second derivatives, to estimate model parameters and their uncertainties from atmospheric observations and their uncertainties. In prognostic mode, we use first derivatives, to map model parameters and their uncertainties onto prognostic quantities and their uncertainties.
P. Cusdin and J.-D. Müller. Improving the performance of code generated by automatic differentiation. Technical Report QUB-SAE-03-04, QUB School of Aeronautical Engineering, 2003.
R. Giering and T. Kaminski. Recomputations in reverse mode AD. In George Corliss, Andreas Griewank, Christele Fauré, Laurent Hascoet, and Uwe Naumann, editors, Automatic Differentiation of Algorithms: From Simulation to Optimization, chapter 33, pages 283-291. Springer Verlag, Heidelberg, 2002. [ http | .pdf ]
R. Giering and T. Kaminski. On the performance of derivative code generated by TAMC. submitted to Optimization Methods and Software, 2002. [ .ps.gz | .pdf ]
W. Mühlhuber. Efficient Solvers for Optimal Design Problems with PDE Constraints. PhD thesis, Johannes Kepler University Linz, 2002. [ http | .ps.gz ]
C. H. Bischof, H. M. Bücker, B. Lang, and A. Rasch. Recent Progress in Automatic Differentiation: Advanced Tools and Large-Scale Applications. Preprint RWTH-CS-SC-01-18, RWTH, Institute for Scientific Computing, Aachen, Germany, 2001. [ .html ]
Among other things, the paper compares the performance of two sets of derivative code for a test problem. One of these sets is generated by ADIFOR 3 and the other by TAF. The TAF generated adjoint code is by a factor of 5 faster than that generated by the ADIFOR 3.
Mark S. Gockenbach. A primer on differentiation. Optimization and Engineering, 2(1):75-129, 2001.
R. Giering. Tangent linear and adjoint biogeochemical models. In Prasad S. Kasibhatla, editor, Inverse Methods in Global Biogeochemical Cycles, volume 114, pages 33-48. American Geophysical Union, 2000. [ .ps | .pdf ]
M. S. Gockenbach. Understanding code generated by TAMC. IAAA Paper TR00-29, Department of Computational and Applied Mathematics, Rice University, Texas, USA, 2000. [ http | .ps ]
Mark uses a number of examples to describe and analyse in detail how TAMC works.
P. Rayner, R. Giering, T. Kaminski, R. Ménard, R. Todling, and C. Trudinger. Exercises. In P. Kasibhatla, M. Heimann, D. Hartley, P. J. Rayner, N. Mahowald, and R. Prinn, editors, Inverse Methods in Global Biogeochemical Cycles, Geophys. Monogr. Ser., volume 114, pages 324-347. American Geophysical Union, Washington, D. C., 1999. [ .pdf ]
R. Giering and T. Kaminski. Using TAMC to generate efficient adjoint code: Comparison of automatically generated code for evaluation of first and second order derivatives to hand written code from the minpack-2 collection. In C. Faure, editor, Automatic Differentiation for Adjoint Code Generation, pages 31-37. INRIA, Sophia Antipolis, France, 1998. [ .html | .ps.gz | .pdf ]
The Paper presents performance numbers of Adjoints and Hessian times Vector codes for functions from the Minpack-2 collection
R. Giering and T. Kaminski. Recipes for Adjoint Code Construction. ACM Trans. Math. Software, 24(4):437-474, 1998. [ DOI | .ps.gz | .pdf ]
R. Giering and T. Kaminski. Recipes for Adjoint Code Construction. Technical Report 212, Max-Planck-Institut für Meteorologie, 1996. [ .html ]
R. Giering. AMC: Ein Werkzeug zum automatischen Differenzieren von Fortran-Programmen. In Theo Plesser and Peter Wittenburg, editors, Forschung und wissenschaftliches Rechnen, Beiträge zum Heinz-Billing-Preis 1995, volume 42, pages 11-27, Göttingen, Germany, 1996. Gesellschaft für wissenschaftliche Datenverarbeitung mbH. [ .html ]
Ralf was awarded the Heinz Billing Price for scientific computing for AMC, the grand father of TAF.
R. Giering. Adjoint code generation. In C.H. Bischof, A. Griewank, and P.M. Khademi, editors, Workshop Report on First Theory Institut on Computational Differentiation, pages 11-12. Technical Report ANL/MCS-TM-183, 1993. [ http ]