This variant on the Cholesky decomposition avoid the need to take the square root by performing the following decomposition:
L*D*LT=A
where L is a lower triangular matrix with zeros on the diagonal. D is a diagonal matrix. The diagonal elements of L are equal to one.
Unfortunately the speed advantage of not computing the square root is washed out by the increased number of array accesses. There only appears to be a slight speed boost for very small matrices.
@author Peter Abeles
|
|
|
|
|
|