Publication: DiffSharp: Automatic Differentiation Library
Introduction
Applications
Tools
Research Groups
Workshops
Publications
   List Publications
   Advanced Search
   Info
   Add Publications
My Account
About

DiffSharp: Automatic Differentiation Library

- Technical report -
 

Author(s)
Atilim Gunes Baydin , Barak A. Pearlmutter , Jeffrey Mark Siskind

Institution
arXiv

Year
2015

Abstract
In this paper we introduce DiffSharp, an automatic differentiation (ad) library designed with machine learning in mind. ad is a family of techniques that evaluate derivatives at machine precision with only a small constant factor of overhead, by systematically applying the chain rule of calculus at the elementary operator level. DiffSharp aims to make an extensive array of ad techniques available, in convenient form, to the machine learning community. These including arbitrary nesting of forward/reverse ad operations, ad with linear algebra primitives, and a functional API that emphasizes the use of higher-order functions and composition. The library exposes this functionality through an API that provides gradients, Hessians, Jacobians, directional derivatives, and matrix-free Hessian- and Jacobian-vector products. Bearing the performance requirements of the latest machine learning techniques in mind, the underlying computations are run through a high-performance BLAS/LAPACK backend, using OpenBLAS by default. GPU support is currently being implemented.

AD Tools
DiffSharp

AD Theory and Techniques
Implementation Strategies

BibTeX
@TECHREPORT{
         Baydin2015DAD,
       title = "DiffSharp: Automatic Differentiation Library",
       author = "Atilim Gunes Baydin and Barak A. Pearlmutter and Jeffrey Mark Siskind",
       year = "2015",
       number = "arXiv:1511.07727",
       institution = "arXiv",
       abstract = "In this paper we introduce DiffSharp, an automatic differentiation (AD) library
         designed with machine learning in mind. AD is a family of techniques that evaluate derivatives at
         machine precision with only a small constant factor of overhead, by systematically applying the
         chain rule of calculus at the elementary operator level. DiffSharp aims to make an extensive array
         of AD techniques available, in convenient form, to the machine learning community. These including
         arbitrary nesting of forward/reverse AD operations, AD with linear algebra primitives, and a
         functional API that emphasizes the use of higher-order functions and composition. The library
         exposes this functionality through an API that provides gradients, Hessians, Jacobians, directional
         derivatives, and matrix-free Hessian- and Jacobian-vector products. Bearing the performance
         requirements of the latest machine learning techniques in mind, the underlying computations are run
         through a high-performance BLAS/LAPACK backend, using OpenBLAS by default. GPU support is currently
         being implemented.",
       ad_tools = "DiffSharp",
       ad_theotech = "Implementation Strategies"
}


back
  

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