wegamekinglc / Derivatives-Algorithms-Lib / 36131645535

push

github

feat: add retrained-policy LSMC AAD sensitivities (#432)

* Add retrained-policy LSMC sensitivities

Expose an opt-in common-path policy secant alongside the existing frozen-policy AAD derivative. Refit model and script-constant bumps on the same disjoint training stream, replay their policies without extra tape recordings, and document the estimator and cross-backend validation.

* Reduce policy sensitivity analysis complexity

* Address LSMC policy risk review feedback

* Keep Excel simulation setting dispatch simple

* Keep generated Excel settings help within registration limit

137 of 139 new or added lines in 3 files covered. (98.56%)

16013 of 17397 relevant lines covered (92.04%)

3333975.0 hits per line