Publication: Accurate Numerical Derivatives in Matlab
Introduction
Applications
Tools
Research Groups
Workshops
Publications
   List Publications
   Advanced Search
   Info
   Add Publications
My Account
About

Accurate Numerical Derivatives in Matlab

- Article in a journal -
 

Author(s)
L. F. Shampine

Published in
ACM Transactions on Mathematical Software

Year
2007

Abstract
Complex step differentiation (CSD) is a technique for computing very accurate numerical derivatives in languages that support complex arithmetic. We describe here the development of a CSD package in Matlab called PMAD. We have extended work done in other languages for scalars to the arrays that are fundamental to Matlab. This extension raises questions that we have been able to resolve in a satisfactory way. Our goal has been to make it as easy as possible to compute approximate Jacobians in Matlab that are all but exact. Although PMAD has a fast option for the expert that implements CSD as in previous work, the default is an object-oriented implementation that asks very little of the user.

AD Theory and Techniques
Complex Step Differentiation

BibTeX
@ARTICLE{
         Shampine2007AND,
       author = "L. F. Shampine",
       title = "Accurate Numerical Derivatives in {Matlab}",
       journal = "{ACM} Transactions on Mathematical Software",
       volume = "33",
       number = "4",
       year = "2007",
       pages = "26",
       note = "Article 26, 17 pages",
       URL = "http://doi.acm.org/10.1145/1268776.1268781",
       abstract = "Complex step differentiation (CSD) is a technique for computing very accurate
         numerical derivatives in languages that support complex arithmetic. We describe here the development
         of a CSD package in Matlab called PMAD. We have extended work done in other languages for scalars to
         the arrays that are fundamental to Matlab. This extension raises questions that we have been able to
         resolve in a satisfactory way. Our goal has been to make it as easy as possible to compute
         approximate Jacobians in Matlab that are all but exact. Although PMAD has a fast option for the
         expert that implements CSD as in previous work, the default is an object-oriented implementation
         that asks very little of the user.",
       month = "aug",
       ad_theotech = "Complex Step Differentiation"
}


back
  

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