Network Embeddings

Statistical analysis of network data.

Inputs

• Network: An instance of Network Graph.

Outputs

• Embeddings: A table with node embeddings

Network Embeddings computes node embeddings using Node2Vec.

Random walk parameters

• Return parameter sets the likelihood of returning to the previous node.

• In-out parameter sets the likelihood of exploring outward nodes.

• Walk length sets the length of each random walk.

• Walks per node sets the number of random walks to start at each node.

Embedding parameters

• Embedding dimension sets the number of dimensions for the embedding space.

• Window size sets the size of the context window for the Skip-gram model.

• Number of epochs sets the number of training iterations for the Skip-gram model.

Example

We loaded the network data from the Dicty publication network from demonstration networks and computed node embeddings. We used the k-means clustering to cluster the node according to their embeddings.

We then plotted the network in Network Explorer, where we colored the nodes according to their cluster assignment.