refnd

refnd is a Python binding for the refnd Rust crate. It provides fast, SIMD-accelerated tools for working with datasets generated by a Random Generative Process (RGP) — datasets where items share latent structure (e.g. protein families, molecules) that makes naive random splits leak information between train and test.

Modules

• refnd.utils
  • BitFingerprint
  • RealFingerprint
  • read_fasta()
  • largest_cluster()
• refnd.kernels
  • Kernel variant selector
  • refnd.kernels.protein.sequence
  • refnd.kernels.molecules
• refnd.core
  • Graph construction
  • HNSW approximate nearest neighbours
  • Graph algorithms
  • Exact search