Argonne workshop discusses theory and applications of automatic differentiationAugust 28, 2015
Argonne National Laboratory hosted the Third Argonne Workshop on Automatic Differentiation on August 20-21, 2015.
Automatic, or algorithmic, differentiation (AD) is a technique for transforming subprograms that compute some mathematical function into subprograms that compute the derivatives of that function. The resulting derivatives are used for uncertainty quantification, optimization algorithms, nonlinear solvers for discretized differential equations, and solution of inverse problems using nonlinear least squares.
The 1.5-day workshop provided a forum for the presentation of developments in both theory and applications of AD. Participants included researchers from Germany, the United Kingdom, and the United States.
“We were particularly pleased to have a diverse mix of mathematicians and computer scientists as well as early career and established researchers.,” said Sri Hari Krishna Narayanan, an assistant computer scientist in Argonne’s Mathematics and Computer Science Division and co-organizer of the workshop.
Program topics ranged from adjoint computation, to the use of AD with Open MP and MPI, to aerodynamic design and computational fluid dynamics. Discussions included completed work, work in progress, and problem areas for future AD exploration.
The full program is available on the web.