...

Author: asavine

differential.png

@@ -24,7 +24,7 @@ AAD arguably constitutes the most significant progress in computation finance of
 
 *New implementations of AAD are pushing the limits of its efficiency, while quantitative analysts are leveraging them in unexpected ways, besides the evident application to risk sensitivities or calibration.*
 
-To a large extent, differential machine learning is another strong application of AAD. It is AAD that gave us the massive number of accurate differentials necessary to implement it, for a very cheap computation cost, and is ultimately responsible for its spectacular performance improvement. 
+To a large extent, differential machine learning is another strong application of AAD. It is AAD that gave us the massive number of accurate differentials necessary to implement it, for a very cheap computation cost, and is ultimately responsible for its spectacular performance improvement. The real-world examples in the Risk paper, sections 3.2 and 3.3, were trained on AAD differential labels.
 
 The Risk paper or the complements do not cover AAD. Readers are referred to the (stellar) founding paper. [This textbook](https://www.amazon.com/Modern-Computational-Finance-Parallel-Simulations-dp-1119539455/dp/1119539455) provides a complete, up to date overview of AAD, its applications in finance, and a complete, professional implementation in modern C++.