Re: [Quantlib-dev] Longstaff-Schwartz method, SVD and OLS

[Klaus S. <kl...@sp...>]

Hi

On Wednesday 08 September 2010 15:42:04 Kakhkhor Abdijalilov wrote:
> SVD can deal with collinearity, but the cutoff threshold for small
> singular values in LinearLeastSquaresRegression is n*QL_EPSILON (n is
> matrix size).
>
> Btw, the cutoff should be applied to the ratio of singular values, not
> to the singular values directly. That is something we need to fix as
> well.

correct. What do you think about the following bug-fix for OLS purposes

change line 116 of linearleastsquaresregression.hpp towards

        const Real threshold 
            = n*QL_EPSILON*(*std::max_element(w.begin(), w.end()));



best regards 
 Klaus