Adept 2.0: a combined automatic differentiation and array library for C++ (original) (raw)
Published October 3, 2017 | Version v1
Software Open
- 1. ECMWF
Description
Adept (Automatic Differentiation using Expression Templates) is a free C++ software library that enables algorithms to be automatically differentiated, very useful for a wide range of applications that involve mathematical optimization. It uses an operator overloading approach, so very little code modification is required. Moreover, the way that expression templates have been used and several other important optimizations mean that reverse-mode differentiation is significantly faster than most other C++ libraries that provide equivalent functionality.
The fundamental object of most mathematical software is the array, and it is therefore natural to want to differentiate array expressions. There are numerous mature C++ array libraries that support a wide variety of array operations, but because they also use expression templates to optimize array expressions, they do not work efficiently with Adept. The solution provided in version 2 of Adept is to use a single expression-template framework that incorporates both array and automatic-differentiation capabilities at a low level, thus enabling efficient differentiation of array expressions. Arrays of up to 7 dimensions are supported along with matrix multiplication and basic linear algebra operations. The library also works well with passive arrays (those that do not need to be differentiated).
Files
Files (653.7 kB)