Sparse Cholesky Elimination Tree
Here I derive the elimination tree for the (right-looking) sparse Cholesky algorithm for computing A = LL^T
for lower triangular L
and sparse matrices A
. This tree forms the foundation for most sparse factorization software, even when the underlying assumptions of Cholesky (symmetric and definite) do not apply. Ultimately this tree tells us two things: 1. where nonzeros appear in the matrix L
even if not present in the original A
(i.e. “fill-in”) and 2. the task dependency graph of our...
