(Kassel, January 26, 1950 – Trento, September 16, 2021) was a German mathematician, known for his extensive contributions to the field of Automatic Differentiation (AD).
After graduating from high school in 1968, he studied in Hofgeismar, at the Clausthal University of Technology and the Albert Ludwig University of Freiburg, where he earned his diploma for work with Lutz Eichner on affine-linear automata. During his postgraduate studies, he completed a master’s degree in 1977 at the University of Canberra, and obtained his PhD in 1980 under the supervision of Richard P. Brent, for his research on “Analysis and Modification of Newton’s Method at Singularities.”
He held a postdoctoral position under Michael J. D. Powell at the University of Cambridge. In 1982 he became an Assistant Professor, and in 1986 an Associate Professor (tenured) at Southern Methodist University in Dallas. From 1987 onward, he worked as a mathematician at Argonne National Laboratory, and from 1992 as a Senior Mathematician. His early research focused on quasi-Newton methods, and from 1987 at Argonne he worked on automatic differentiation. In 1988 he wrote the program ADOL-C for this purpose. In 1998/99 he was at INRIA Sophia Antipolis in Antibes. In 2001, he received a research award from the Max Planck Society in mathematics and computer science.
In 1993, Griewank became professor and director of the Institute for Scientific Computing at the Dresden University of Technology. From 2003 to 2015 he was professor at Humboldt University of Berlin, and from 2008 he served as director of the Institute of Mathematics.
Andreas Griewank’s research spanned a very broad area of mathematical optimization. It included, for example, the convergence theory of Newton’s method in the degenerate infinite-dimensional case; approaches to global and non-uniform optimization; and the efficient computation of exact derivatives via algorithmic differentiation.
Andreas is widely regarded as the godfather of Automatic Differentiation (AD). He began working in this research area in the early 1980s and quickly became a driving force behind AD’s development and one of the field’s leading proponents.
The so-called Griewank function, which serves as an academic test function in the field of global optimization, is another of Andreas’s contributions. It is widely used in the global optimization community and has attracted renewed interest as non-convex optimization underpins objective minimization in data-analysis applications such as deep learning.
