I just realized that the attachment exceeds (by far) the maximum size. With apologies (this is my first post), re-posting now without the attachment. The introduction can be found at the top level of the repository, under the name AADIntro.pdf Thank you. Antoine Savine From: Antoine Savine Sent: 22 November 2017 12:38 To: 'boost@lists.boost.org' <boost@lists.boost.org> Subject: Interest for Algorithmic Adjoint Differentiation (AAD) in Boost? AAD is a hot topic in machine learning, computational finance, and, I think, other disciplines such as meteorology. Adjoint differentiation computes a large number of derivatives sensitivities for some calculation code in constant time, and operator overloading in C++ permits its semi-automatic implementation over templated code. Please see the intro attached for details. I have been working with AAD (in finance) for 5 years. I have been developing and using professional AAD libraries, I have given dedicated talks and workshops at professional conferences such as Global Derivatives and WBS, and I have been teaching AAD at the MsC in Mathematical Finance at the Copenhagen University in Denmark (Mathematics Department). Our implementation in Danske Bank obtained in 2015 the In-House System of the Year 2015 Risk award. I am planning a book on AAD. And, although that publication is biased towards finance, I have developed an independent, simple and efficient AAD engine that can be found here: tinyUrl.com/aadCentral The implementation uses the latest developments in AAD technology for maximum performance, in particular expression templates. I thought that given the wide interest for AAD, a Boost.Aad module could make sense and I was hoping that it could be based on my developments. Please could somebody let me know if there is interest, and, if so, how I should proceed? Million thanks. Kind regards, Antoine Savine