Edge Feature

gtrick.pyg.CommonNeighbors(edge_index, edge_attr=None, batch_size=64)

Bases: object

Compute the common neighbors of two nodes in a graph.

Example

EdgeFeat (PyG)

Parameters:

Name	Type	Description	Default
edge_index	torch.LongTensor	Graph connectivity.	required
edge_attr	torch.Tensor	Edge feature.	None
batch_size	int	The batch size to compute common neighbors.	64

__call__(edges)

Parameters:

Name	Type	Description	Default
edges	torch.Tensor	The edges with the shape (num_edges, 2).	required

Returns:

Type	Description
torch.Tensor	The calculated common neighbors feature.

gtrick.pyg.ResourceAllocation(edge_index, edge_attr=None, batch_size=64)

Bases: object

Compute the resource allocation of two nodes in a graph.

Resource allocation of \(u\) and \(v\) is defined as

\[ \sum_{w \in \Gamma(u) \cap \Gamma(v)} \frac{1}{|\Gamma(w)|} \]

where \(\Gamma(u)\) denotes the set of neighbors of \(u\).

Example

EdgeFeat (PyG)

Parameters:

Name	Type	Description	Default
edge_index	torch.LongTensor	Graph connectivity.	required
edge_attr	torch.Tensor	Edge feature.	None
batch_size	int	The batch size to compute common neighbors.	64

__call__(edges)

Parameters:

Name	Type	Description	Default
edges	torch.Tensor	The edges with the shape (num_edges, 2).	required

Returns:

Type	Description
torch.Tensor	The calculated resource allocation feature.

gtrick.pyg.AdamicAdar(edge_index, edge_attr=None, batch_size=64)

Bases: object

Computes the adamic adar of two nodes in a graph.

Adamic-Adar index of \(u\) and \(v\) is defined as

\[ \sum_{w \in \Gamma(u) \cap \Gamma(v)} \frac{1}{\log |\Gamma(w)|} \]

where \(\Gamma(u)\) denotes the set of neighbors of \(u\). This index leads to zero-division for nodes only connected via self-loops. It is intended to be used when no self-loops are present.

Example

EdgeFeat (PyG)

Parameters:

Name	Type	Description	Default
edge_index	torch.LongTensor	Graph connectivity.	required
edge_attr	torch.Tensor	Edge feature.	None
batch_size	int	The batch size to compute common neighbors.	64

__call__(edges=None)

Parameters:

Name	Type	Description	Default
edges	torch.Tensor	The edges with the shape (num_edges, 2).	None

Returns:

Type	Description
torch.Tensor	The calculated adamic adar feature.

gtrick.pyg.AnchorDistance(data, num_samples, k, ka, max_spd=5, to_undirected=True)

Bases: object

Computes the anchor distance of two nodes in a graph.

The anchor distance randomly selects \(k_a\) nodes from \(V\) to be anchor nodes and then calculates the shortest path starting from these anchor nodes to any other nodes. After that, the distance between \(u\) and \(v\) can be estimated by:

\[ d_{u, v}=\frac{1}{K_{A}} \sum_{i=1}^{K_{A}} d_{u, a_{i}}+d_{v, a_{i}} \]

To reduce the randomness, it uses \(k\) anchor sets to generate multiple distance features.

Example

EdgeFeat (PyG)

Parameters:

Name	Type	Description	Default
data	torch_geometric.data.Data	Graph data.	required
num_samples	int	The number of times to sample anchor sets.	required
k	int	The size of sampled anchor sets.	required
ka	int	The number of anchor nodes.	required
max_spd	int	The max shortest distance.	5
to_undirected	bool	Converts the graph to an undirected graph.	True

__call__(edges)

Parameters:

Name	Type	Description	Default
edges	torch.Tensor	The edges with the shape (num_edges, 2).	required

Returns:

Type	Description
torch.Tensor	The calculated anchor distance feature.