Publication: Automatic differentiation for error analysis of Monte Carlo data
Introduction
Applications
Tools
Research Groups
Workshops
Publications
   List Publications
   Advanced Search
   Info
   Add Publications
My Account
About

Automatic differentiation for error analysis of Monte Carlo data

- Article in a journal -
 

Area
Error Analysis

Author(s)
Alberto Ramos

Published in
Computer Physics Communications

Year
2019

Abstract
Automatic Differentiation (ad) allows to determine exactly the Taylor series of any function truncated at any order. Here we propose to use ad techniques for Monte Carlo data analysis. We discuss how to estimate errors of a general function of measured observables in different Monte Carlo simulations. Our proposal combines the Γ-method with Automatic differentiation, allowing exact error propagation in arbitrary observables, even those defined via iterative algorithms. The case of special interest where we estimate the error in fit parameters is discussed in detail. We also present a freely available fortran reference implementation of the ideas discussed in this work.

AD Tools
aderrors

BibTeX
@ARTICLE{
         Ramos2019Adf,
       title = "Automatic differentiation for error analysis of {M}onte {C}arlo data",
       journal = "Computer Physics Communications",
       volume = "238",
       pages = "19--35",
       year = "2019",
       issn = "0010-4655",
       doi = "10.1016/j.cpc.2018.12.020",
       url = "https://www.sciencedirect.com/science/article/pii/S0010465519300013",
       author = "Alberto Ramos",
       keywords = "Lattice QCD, Monte Carlo, Error analysis",
       abstract = "Automatic Differentiation (AD) allows to determine exactly the Taylor series of any
         function truncated at any order. Here we propose to use AD techniques for Monte Carlo data analysis.
         We discuss how to estimate errors of a general function of measured observables in different Monte
         Carlo simulations. Our proposal combines the Γ-method with Automatic differentiation,
         allowing exact error propagation in arbitrary observables, even those defined via iterative
         algorithms. The case of special interest where we estimate the error in fit parameters is discussed
         in detail. We also present a freely available fortran reference implementation of the ideas
         discussed in this work.",
       ad_area = "Error Analysis",
       ad_tools = "aderrors"
}


back
  

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