graphletStructural graph mining for petgraph

Subgraph census, graphlet-degree vectors (GDV/GDD), per-node orbits, and network-motif detection — one census substrate, minimal dependencies.

Introduction

View on GitHub

One census substrate

enumerate connected k-subsets → canonically label → fold. Instance enumeration and counting are two readouts of a single pass; counting streams, so memory tracks graph size, not the number of instances.

Orbits, GDV & GDD

Per-node attribution to all 73 automorphism orbits (k ≤ 5, undirected), verified against an independent brute-force oracle.

Named motifs, honest semantics

Induced and non-induced (monomorphism) counts for catalog patterns — the non-induced readout derived from the induced census via a fixed s(P,C) table, no separate enumerator.

Minimal dependencies

Depends only on petgraph and rand. Generic over Graph / StableGraph, directed / undirected, and arbitrary weights through one trait-bound set.