Efficient Calculation of Jacobians Using Dynamic Programming - Archive ouverte HAL Access content directly
Reports Year : 1999

Efficient Calculation of Jacobians Using Dynamic Programming


The chain rule - fundamental for Automatic Differentiation (AD) - can be applied to computational graphs representing vector functions in arbitrary orders resulting in different operations counts for the calculation of their Jacobian matrices. Very few authors have looked at this interesting subject so far and there is no generally accepted terminology for dealing with these combinations of the forward and reverse modes of AD. The minimizati- on of the number of arithmetic operations required for the calculation of the complete Jacobian leads to a computationally hard combinatorial optimization problem. In this paper we will describe an approach to the solution of the edge elimination problem in computational graphs that builds on the idea of optimizing chained matrix products using dynamic programming techniques. We will discuss the theory and present some test results by regarding this approach in comparison with other methods for computing Jacobians using a minimal number of arithmetic operations.
Fichier principal
Vignette du fichier
RR-3689.pdf (537.69 Ko) Télécharger le fichier

Dates and versions

inria-00072980 , version 1 (24-05-2006)


  • HAL Id : inria-00072980 , version 1


Uwe Naumann. Efficient Calculation of Jacobians Using Dynamic Programming. RR-3689, INRIA. 1999. ⟨inria-00072980⟩
62 View
336 Download


Gmail Facebook Twitter LinkedIn More