fastnode2vec 0.0.3

Fast implementation of node2vec

fastnode2vec

Really fast implementation of node2vec based on numba and gensim.

API

Node2Vec inherits from gensim's Word2Vec, all its APi is valid.

from fastnode2vec import Graph, Node2Vec

graph = Graph([("a", "b"), ("b", "c"), ("c", "a"), ("a", "d")],
              directed=False, weighted=False)

# or
graph = Graph([("a", "b", 1), ("b", "c", 2), ("c", "a", 3), ("a", "d", 4)],
              directed=False, weighted=True)

n2v = Node2Vec(graph, dim=10, walk_length=100, context=10, p=2.0, q=0.5, workers=2)

n2v.train(epochs=100)

print(n2v.wv["a"])

CLI

Usage: fastnode2vec [OPTIONS] FILENAME

Options:
  --directed
  --dim INTEGER          [required]
  --p FLOAT
  --q FLOAT
  --walk-length INTEGER  [required]
  --context INTEGER
  --epochs INTEGER       [required]
  --workers INTEGER
  --batch-walks INTEGER
  --debug PATH
  --output PATH
  --help                 Show this message and exit.

Compute embeddings of the Gnutella peer-to-peer network:

wget https://snap.stanford.edu/data/p2p-Gnutella08.txt.gz
fastnode2vec p2p-Gnutella08.txt.gz --dim 16 --walk-length 100 --epochs 10 --workers 2

Load embeddings produced by the CLI

Just use the Word2Vec API.

from gensim.models import KeyedVectors

wv = KeyedVectors.load("p2p-Gnutella08.txt.gz.wv", mmap='r')