Making automatic differentiation truly automatic: coupling PETSc with ADIC
نویسندگان
چکیده
منابع مشابه
Parallel simulation of compressible flow using automatic differentiation and PETSc
Many aerospace applications require parallel implicit solution strategies and software. We consider the use of two computational tools, PETSc and ADIFOR, to implement a Newton-Krylov-Schwarz method with pseudo-transient continuation for a particular application, namely, a steady-state, fully implicit, three-dimensional compressible Euler model of ow over an M6 wing. We describe how automatic di...
متن کاملADIC: An Extensible Automatic Differentiation Tool for ANSI-C
In scienti c computing, we often require the derivatives @f=@x of a function f expressed as a program with respect to some input parameter(s) x, say. Automatic di erentiation (AD) techniques augment the program with derivative computation by applying the chain rule of calculus to elementary operations in an automated fashion. This article introduces ADIC (Automatic Di erentiation of C), a new A...
متن کاملAutomatic Differentiation with Code Coupling and Applications to Scale Modules
An advanced automatic differentiation tool for Fortran 90 software has been developed at Oak Ridge National Laboratory. This tool, called GRESS 90, has a code-coupling feature to propagate derivatives relative to the input of one code through a series of codes that utilize the results of one calculation as the input in the next to determine a final result. GRESS 90 has been applied to the reson...
متن کاملAutomatic differentiation in PyTorch
In this article, we describe an automatic differentiation module of PyTorch — a library designed to enable rapid research on machine learning models. It builds upon a few projects, most notably Lua Torch, Chainer, and HIPS Autograd [4], and provides a high performance environment with easy access to automatic differentiation of models executed on different devices (CPU and GPU). To make prototy...
متن کاملOn Automatic Differentiation
In comparison to symbolic diierentiation and numerical diierencing, the chain rule based technique of automatic diierentiation is shown to evaluate partial derivatives accurately and cheaply. In particular it is demonstrated that the reverse mode of automatic diierentiation yields any gradient vector at no more than ve times the cost of evaluating the underlying scalar function. After developin...
متن کاملذخیره در منابع من
با ذخیره ی این منبع در منابع من، دسترسی به آن را برای استفاده های بعدی آسان تر کنید
ژورنال
عنوان ژورنال: Future Generation Computer Systems
سال: 2005
ISSN: 0167-739X
DOI: 10.1016/j.future.2004.11.008