Adjoint Venture: Fast Greeks with Adjoint Algorithmic Differentiation
This dissertation seeks to discuss the adjoint approach to solving affine recursion problems (ARPs) in the context of computing sensitivities of financial instruments. It is shown how, by moving from an intuitive 'forward' approach to solving a recursion to an 'adjoint' approach,...
| Main Author: | |
|---|---|
| Other Authors: | |
| Format: | Dissertation |
| Language: | English |
| Published: |
University of Cape Town
2017
|
| Subjects: | |
| Online Access: | http://hdl.handle.net/11427/25412 |
| Summary: | This dissertation seeks to discuss the adjoint approach to solving affine recursion problems (ARPs) in the context of computing sensitivities of financial instruments. It is shown how, by moving from an intuitive 'forward' approach to solving a recursion to an 'adjoint' approach, one might dramatically increase the computational efficiency of algorithms employed to compute sensitivities via the pathwise derivatives approach in a Monte Carlo setting. Examples are illustrated within the context of the Libor Market Model. Furthermore, these ideas are extended to the paradigm of Adjoint Algorithmic Differentiation, and it is illustrated how the use of sophisticated techniques within this space can further improve the ease of use and efficiency of sensitivity calculations. |
|---|