Graph Embeddings

Graph Embeddings are a way to learn a low dimensional representation of a graph. Through a graph embedding a graph can be represented in low dimensional form, while preserving graph information. This low-dimensional representation can then be used for training classic machine learning algorithms that would otherwise make no use of the graph information.

The Graph Embeddings used by PHOTON Graph are static graph embeddings, based on the gem python package.

GraphEmbeddingHOPE

Transformer class for calculating a Graph Embedding based on Higher-order proximity preserved embedding (Mingdong et al., 2016). Implementation based on gem python package.

Parameters:

Name	Type	Description	Default
embedding_dimension	int	the number of dimensions that the final embedding will have	1
decay_factor	float	the higher order coefficient beta	0.01
adjacency_axis	int	position of the adjacency matrix, default being zero	0
logs	str	Path to the log data	None

Example

Text Only

constructor = GraphEmbeddingHOPE(embedding_dimension=1,
                                 decay_factor=0.1)

GraphEmbeddingLaplacianEigenmaps

Transformer class for calculating a Graph Embedding based on Laplacian Eigenmaps (Belkin & Niyogi, 2013). Implementation based on gem python package.

Parameters:

Name	Type	Description	Default
embedding_dimension	int	the number of dimensions that the final embedding will have	1
adjacency_axis	int	position of the adjacency matrix, default being zero	0
logs	str	Path to the log data	None

Example

Python

constructor = GraphEmbeddingLaplacianEigenmaps(embedding_dimension=1)

GraphEmbeddingLocallyLinearEmbedding

Transformer class for calculating a Graph Embedding based on Locally Linear Embedding (Roweis & Saul, 2000). Implementation based on gem python package.

Parameters:

Name	Type	Description	Default
embedding_dimension	int	the number of dimensions that the final embedding will have	1
adjacency_axis	int	position of the adjacency matrix, default being zero	0
logs	str	Path to the log data	None

Example

Python

constructor = GraphEmbeddingLocallyLinearEmbedding(embedding_dimension=1)