NBNE

Code to use Neighbor Based Node Embeddings (NBNE) method to create representations to nodes in a graph.

Installation

You can install NBNE directly from PyPI:

pip install nbne

Or from source:

git clone https://github.com/tiagopms/nbne.git
cd nbne
pip install .

Dependencies

NBNE has the following requirements:

• NetworkX
• Gensim

Usage

Basic Usage

The libraries gensim and networkx should be installed. Then run:

    $ nbne --input examples/data/watts_strogatz.graph --output examples/data/watts_strogatz.emb

Using in other Applications

Import nbne module in your application and train model with:

    from nbne import train_model
    train_model(graph, num_permutations)

Where graph should be a networkx graph. To save the model in an output file:

    from nbne import train_model
    import networkx as nx
    graph = nx.watts_strogatz_graph(1000, 50, 0.2)
    train_model(graph, num_permutations, output_name)

Input

Input should be a edgelist with format:

    node1_id node2_id
    node1_id node3_id
    node2_id node3_id

Output

The output is a document with n+1 lines. The first has format:

    num_nodes embeddings_size

And the other:

    node_id embedding

Where embedding is a space separated vector with dimension d, i.e. d1 d2 d3 ... dn.