num::randommat::act

Defined in header <linear/graph/randommat/approxchol.hpp>

template <typename Float = double, std::integral Index = num::idx, typename Rng = rng64>
cholesky_factor<Float, Index> act(const adjacency_list<Float, Index> &G, Rng &rng,
                                  std::type_identity_t<Index> trees = 1);

Spanning-tree clique sampler: one reweighted random tree per elimination.

Keeps \(d-1\) edges per eliminated vertex instead of \(d\), drawn from a negatively dependent distribution rather than independently. See num::structures::sample_clique_tree for why the reweighting needs no effective-resistance solve.

Return value

cholesky_factor: the sparse approximate factor, consumed by randommat::solve.