Publication: Automatic adjoint differentiation for gradient descent and model calibration
Introduction
Applications
Tools
Research Groups
Workshops
Publications
   List Publications
   Advanced Search
   Info
   Add Publications
My Account
About

Automatic adjoint differentiation for gradient descent and model calibration

- Article in a journal -
 

Author(s)
Dmitri Goloubentsev , Evgeny Lakshtanov

Published in
International Journal of Wavelets, Multiresolution and Information Processing

Year
2020

Abstract
In this work, we discuss the Automatic Adjoint Differentiation (AAD) for functions of the form G=(1) ⁄ (2) ∑_i=1^m (Ey_i-C_i)^2 which often appear in the calibration of stochastic models. We demonstrate that it allows a perfect Single Input Multiple Data (SIMD) parallelization and provides its relative computational cost. In addition, we demonstrate that this theoretical result is in concordance with numerical experiments.

BibTeX
@ARTICLE{
         Goloubentsev2020Aad,
       author = "Goloubentsev, Dmitri and Lakshtanov, Evgeny",
       title = "Automatic adjoint differentiation for gradient descent and model calibration",
       journal = "International Journal of Wavelets, Multiresolution and Information Processing",
       volume = "0",
       number = "0",
       pages = "2040004",
       year = "2020",
       doi = "10.1142/S0219691320400044",
       url = "https://doi.org/10.1142/S0219691320400044",
       eprint = "https://doi.org/10.1142/S0219691320400044",
       abstract = "In this work, we discuss the Automatic Adjoint Differentiation (AAD) for functions
         of the form $G=\frac{1}{2} \sum_{i=1}^m (Ey_i-C_i)^2$ which often appear in the
         calibration of stochastic models. We demonstrate that it allows a perfect Single Input Multiple Data
         (SIMD) parallelization and provides its relative computational cost. In addition, we demonstrate
         that this theoretical result is in concordance with numerical experiments."
}


back
  

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