|
|
|
| e-Pub |
Previous | 
Home
Section: Research Program
Automatic Differentiation and Scientific Computing
Participants : Alain Dervieux, Laurent Hascoet, Bruno Koobus.
-
Glossary
- linearization
In Scientific Computing, the mathematical model often consists of Partial Derivative Equations, that are discretized and then solved by a computer program. Linearization of these equations, or alternatively linearization of the computer program, predict the behavior of the model when small perturbations are applied. This is useful when the perturbations are effectively small, as in acoustics, or when one wants the sensitivity of the system with respect to one parameter, as in optimization.
- adjoint state
Consider a system of Partial Derivative Equations that define some characteristics of a system with respect to some input parameters. Consider one particular scalar characteristic. Its sensitivity, (or gradient) with respect to the input parameters can be defined as the solution of “adjoint” equations, deduced from the original equations through linearization and transposition. The solution of the adjoint equations is known as the adjoint state.
Scientific Computing provides reliable simulations of complex systems. For example it is possible to simulate the 3D air flow around a plane that captures the physical phenomena of shocks and turbulence. Next comes optimization, one degree higher in complexity because it repeatedly simulates and applies optimization steps until an optimum is reached. We focus on gradient-based optimization.
We investigate several approaches to obtain the gradient, between two extremes:
-
One can write an adjoint system of mathematical equations, then discretize it and program it by hand. This is mathematically sound [25] , but very costly in development time. It also does not produce an exact gradient of the discrete function, and this can be a problem if using optimization methods based on descent directions.
-
One can apply adjoint-mode AD (cf 3.1 ) on the program that discretizes and solves the direct system. This gives in fact the adjoint of the discrete function computed by the program. Theoretical results [24] guarantee convergence of these derivatives when the direct program converges. This approach is highly mechanizable, but leads to massive use of storage and may require code transformation by hand [30] , [33] to reduce memory usage.
If for instance the model is steady, one tradeoff can use the iterated states in the direct order [26] . Another tradeoff can use only the fully converged final state. Since tradeoff approaches can be error-prone, we advocate incorporating them into the AD model and into the AD tools.