Symmetrize an experimental disk-backed UMAP membership graph — massive_umap_fuzzy

Externally sorts directed membership pairs in bounded native buffers, applies fuzzy union, and writes a symmetric weighted CSR graph and a content-hashed manifest. The sort needs substantial temporary disk space and is currently CPU-only. Use massive_fuzzy_graph() after a restart.

Usage

massive_umap_fuzzy_graph(
  x,
  output,
  memory_limit = "8GB",
  backend = "cpu",
  checkpoint = FALSE,
  resume = FALSE,
  checkpoint_every = 100L
)

Arguments

x: A result of massive_umap_memberships().
output: New prefix for .offsets.u64, .indices.u32, and .weights.f32 output files.
memory_limit: RAM budget for external sorting, at least 64 MB.
backend: Only "cpu" is currently supported.
checkpoint: Save completed sorted-pair runs for resume.
resume: Continue a matching interrupted graph sort.
checkpoint_every: Save progress after this many input blocks.

Value

An experimental symmetric fastEmbedR_massive_graph descriptor.