[图书][B] Evaluating derivatives: principles and techniques of algorithmic differentiation
A Griewank, A Walther - 2008 - SIAM
The advent of high-speed computers and sophisticated software tools has made the
computation of derivatives for functions defined by evaluation programs both easier and …
computation of derivatives for functions defined by evaluation programs both easier and …
Algorithm 799: revolve: an implementation of checkpointing for the reverse or adjoint mode of computational differentiation
A Griewank, A Walther - ACM Transactions on Mathematical Software …, 2000 - dl.acm.org
In its basic form, the reverse mode of computational differentiation yields the gradient of a
scalar-valued function at a cost that is a small multiple of the computational work needed to …
scalar-valued function at a cost that is a small multiple of the computational work needed to …
A mathematical view of automatic differentiation
A Griewank - Acta Numerica, 2003 - cambridge.org
Automatic, or algorithmic, differentiation addresses the need for the accurate and efficient
calculation of derivative values in scientific computing. To this end procedural programs for …
calculation of derivative values in scientific computing. To this end procedural programs for …
Optimal Jacobian accumulation is NP-complete
U Naumann - Mathematical Programming, 2008 - Springer
We show that the problem of accumulating Jacobian matrices by using a minimal number of
floating-point operations is NP-complete by reduction from Ensemble Computation. The …
floating-point operations is NP-complete by reduction from Ensemble Computation. The …
A discrete adjoint approach for the optimization of unsteady turbulent flows
R Roth, S Ulbrich - Flow, turbulence and combustion, 2013 - Springer
In this paper we present a discrete adjoint approach for the optimization of unsteady,
turbulent flows. While discrete adjoint methods usually rely on the use of the reverse mode …
turbulent flows. While discrete adjoint methods usually rely on the use of the reverse mode …
Optimal multistage algorithm for adjoint computation
We reexamine the work of Stumm and Walther on multistage algorithms for adjoint
computation. We provide an optimal algorithm for this problem when there are two levels of …
computation. We provide an optimal algorithm for this problem when there are two levels of …
Optimal memory-aware backpropagation of deep join networks
O Beaumont, J Herrmann… - … Transactions of the …, 2020 - royalsocietypublishing.org
Deep learning training memory needs can prevent the user from considering large models
and large batch sizes. In this work, we propose to use techniques from memory-aware …
and large batch sizes. In this work, we propose to use techniques from memory-aware …
Checkpointing schemes for adjoint codes: Application to the meteorological model Meso-NH
I Charpentier - SIAM Journal on Scientific Computing, 2001 - SIAM
The adjoint code of a nonlinear computer model calculates gradients along a trajectory that
has to be known at integration time. When the storage of the whole trajectory requires too …
has to be known at integration time. When the storage of the whole trajectory requires too …
[HTML][HTML] DAG reversal is NP-complete
U Naumann - Journal of Discrete Algorithms, 2009 - Elsevier
Runs of numerical computer programs can be visualized as directed acyclic graphs (DAGs).
We consider the problem of restoring the intermediate values computed by such a program …
We consider the problem of restoring the intermediate values computed by such a program …