Publication: Circumventing Storage Limitations in Variational Data Assimilation
Introduction
Applications
Tools
Research Groups
Workshops
Publications
   List Publications
   Advanced Search
   Info
   Add Publications
My Account
About

Circumventing Storage Limitations in Variational Data Assimilation

- Article in a journal -
 

Area
Oceanography

Author(s)
Juan M. Restrepo , Gary K. Leaf , Andreas Griewank

Published in
SIAM Journal on Scientific Computing

Year
1998

Abstract
Trading computing cycles for memory requirements we show that the gradient of a functional can be computed on a finite memory machine exactly. Furthermore, the algorithmic strategy exchanges in an optimal way storage and computational cycles, hence it can be exploited to efficiently compute the desired gradient. The algorithm and the publicly-available code that implements it are illustrated in the context of a typical geophysical data assimilation problem.

AD Tools
Treeverse / Revolve

AD Theory and Techniques
Checkpointing

Related Applications
- Data Assimilation for Ocean Circulation

BibTeX
@ARTICLE{
         Restrepo1998CSL,
       ad_tools = "Treeverse / Revolve",
       ad_theotech = "Checkpointing",
       ad_area = "Oceanography",
       author = "Juan M. Restrepo and Gary K. Leaf and Andreas Griewank",
       title = "Circumventing Storage Limitations in Variational Data Assimilation",
       journal = "SIAM Journal on Scientific Computing",
       volume = "19",
       pages = "1586-1605",
       year = "1998",
       abstract = "Trading computing cycles for memory requirements we show that the gradient of a
         functional can be computed on a finite memory machine exactly. Furthermore, the algorithmic strategy
         exchanges in an optimal way storage and computational cycles, hence it can be exploited to
         efficiently compute the desired gradient. The algorithm and the publicly-available code that
         implements it are illustrated in the context of a typical geophysical data assimilation problem."
}


back
  

Contact:
autodiff.org
Username:
Password:
(lost password)