Graph Conversion functions

The conversion functions are a collection of different handy helper functions for converting your input graphs between different toolboxes.

Popular supported formats are:

• numpy arrays
• networkx graphs
• dgl graphs
• scipy sparse matrices

PHOTONAI Graph will handle necessary conversions in your pipeline automatically.

Graph Conversion Functions

check_dgl(graphs, adjacency_axis=None, feature_axis=None)

Checks the input and converts it to dgl format

Parameters:

Name	Type	Description	Default
graphs		graphs to be converted	required
adjacency_axis		position of the adjacency matrix	None
feature_axis		position of the feature matrix	None

dense_to_dgl(graphs, adjacency_axis=None, feature_axis=None)

Converts dense matrices to dgl graphs

Parameters:

Name	Type	Description	Default
graphs		graphs represented in dense format	required
adjacency_axis		position of the adjacency matrix	None
feature_axis		position of the feature matrix	None

dense_to_igraph(graphs, adjacency_axis=None, feature_axis=None)

Converts dense matrices to igraph Graph objects :param graphs: np.ndarray or np.matrix graphs represented as dense format :param adjacency_axis: int, default=None position of the adjacency axis :param feature_axis: int, default=None position of the feature axis :return: igraph_graphs: Converted list of Igraph objects

dgl_to_dense(graphs, in_fmt='csr')

turns dgl graphs into dense matrices

Parameters:

Name	Type	Description	Default
graphs		list of dgl graphs	required
in_fmt		format of the scipy sparse matrix used in the intermediary step	'csr'

dgl_to_sparse(graphs, fmt='csr')

turns dgl graphs into sparse matricesParameters

Parameters:

Name	Type	Description	Default
graphs		list of dgl graphs	required
fmt		format of the scipy sparse matrix used in the intermediary step	'csr'

sparse_to_dense(graphs, features=None)

convert sparse matrices to numpy array

Parameters:

Name	Type	Description	Default
graphs		a list of scipy sparse matrices or a single sparse matrix	required
features		if a feature matrix or a list of those is specified they are incorporated into the numpy array	None