Integrating AD in Large Quant Libraries

Algorithmic Differentiation Video Series - Part 3

Many financial institutions have developed large quantitative finance analytics libraries over years, reaching over a million lines of code. Integrate AAD into these libraries becomes a real challenge. Many parts of the code base are affected and performance and memory consumption are major concerns.

The third part of the video series gives practical guidelines on how to cope with the complexities involved when integrating AD into large quant libraries. It outlines integration approaches, details techniques to limit the memory requirements of AAD, explains how external libraries can be integrated into the derivatives calculation and shows how to handle non-differentiable functions. It further gives performance tuning advice.

More in this Series:

• Part 1: AD Theory and Manual Implementation
• Part 2: Automatic Algorithmic Differentiation

‍