Sigma.DiffSharp is a modified version of the DiffSharp library for integration with the Sigma machine learning framework with support for ndarrays, multiple CLI backends and more data types. Sigma.DiffSharp is not intended to be new release of the original DiffSharp library, but rather an adaption with a restricted featureset for interopability with the Sigma project.

DiffSharp is a functional automatic differentiation (AD) library implemented in the F# language. It supports C# and the other CLI languages. The library is being developed mainly for research applications in machine learning, by Atılım Güneş Baydin and Barak A. Pearlmutter, within the Brain and Computation Lab, National University of Ireland Maynooth.

Please visit the project website for detailed documentation and examples.

You can come and join the Gitter chat room, if you want to chat with us:

Copyright (c) 2014–2016, National University of Ireland Maynooth (Atilim Gunes Baydin, Barak A. Pearlmutter)

Written by: Atilim Gunes Baydin & Barak A. Pearlmutter

Brain and Computation Lab
Hamilton Institute & Department of Computer Science
National University of Ireland Maynooth
Maynooth, Co. Kildare
Ireland

www.bcl.hamilton.ie

DiffSharp is released under the GNU Lesser General Public License (LGPL) version 3. This means that you can integrate DiffSharp as a shared library into your software, which may be commercial, closed-source, or open-source under any license (including non-GPL), provided that you use the unmodified DiffSharp binary and reproduce the above copyright notice with a link to the DiffSharp website. Permitting uses outside of these license terms may be considered on a by-case basis.

This work is supported by Science Foundation Ireland grant 09/IN.1/I2637.

DiffSharp uses:

• OpenBLAS by Zhang Xianyi, Wang Qian, Werner Saar (BSD license) for BLAS/LAPACK operations
• F# Quotations Evaluator by Paul Westcott and others (Unlicense/public domain) for compiling code quotations