Vector AAD for Large Portfolios: Applications to XVA, FRTB, Dynamic IM, and KVA/MVA*

The latest research presented by Alexander Sokol at the QuantTech Conference in London (April 21st, 2016) offers a comprehensive overview of implementing Vector AAD for calculations involving large portfolios. In this he demonstrates the mechanics of the performance boost achieved by selecting Reverse Mode AD (also called Adjoint Algorithmic Differentiation (AAD)) over Forward Mode AD.

 

The presentation has been put together for all members of the financial community and explains the fundamental concepts, practical benefits and limitations for overcoming the challenges that naturally arise when dealing with large portfolios, excessive regulatory demands and a limited operational budget.

 

Download the entire presentation here to learn why Adjoint Algorithmic Differentiation has the quant community talking and find out how AAD can be employed in your day-to-day operations for calculating XVA, FRTB, KVA/MVA and other sophisticated computations.

 

 

* Code samples used in the presentation are based on TapeScript, an open source library for Vector AAD available from github.com/compatibl