Expand description
Edge sampling proportional to importance scores.
Implements the spectral sampling strategy: each edge is kept independently
with probability proportional to weight * importance * log(n) / epsilon^2,
and kept edges are reweighted by the inverse of their sampling probability
to preserve the expected Laplacian.
Structsยง
- Spectral
Sampler - Samples edges from a graph proportional to their spectral importance.