cholesky bug

[tschm]

if A is a symmetric positive definite matrix we can compute the Cholesky factorization

A = L * L'

where L is a lower triangular matrix. Funny enough there is a slightly different convention used in scipy vs. dumpy.

Hence the volatility is

sqrt(x' A x) = 2-Norm (L' x)

Note that

2-Norm (L'x) is not equal 2-Norm(Lx)