graphs

A module containing different graph representations in GraphNeT.

class graphnet.models.data_representation.graphs.graphs.KNNGraph(*args, **kwargs)

Bases: GraphDefinition

A Graph representation where Edges are drawn to nearest neighbours.

Construct k-nn graph representation.

Parameters:

• detector (Detector) – Detector that represents your data.

• node_definition (Optional[NodeDefinition], default: None) – Definition of nodes in the graph.

• input_feature_names (Optional[List[str]], default: None) – Name of input feature columns.

• dtype (Optional[dtype], default: torch.float32) – data type for node features.

• perturbation_dict (Optional[Dict[str, float]], default: None) – Dictionary mapping a feature name to a standard deviation according to which the values for this feature should be randomly perturbed. Defaults to None.

• seed (Union[int, Generator, None], default: None) – seed or Generator used to randomly sample perturbations. Defaults to None.

• nb_nearest_neighbours (int, default: 8) – Number of edges for each node. Defaults to 8.

• columns (List[int], default: [0, 1, 2]) – node feature columns used for distance calculation. Defaults to [0, 1, 2].

• distance_as_edge_feature (bool, default: False) – Add edge distances as an edge feature. Defaults to False.

• args (Any)

• kwargs (Any)

Return type:

object

class graphnet.models.data_representation.graphs.graphs.EdgelessGraph(*args, **kwargs)

Bases: GraphDefinition

A Data representation without edge assignment.

I.e the resulting representation is created without an EdgeDefinition.

Construct isolated nodes graph representation.

Parameters:

• detector (Detector) – Detector that represents your data.

• node_definition (Optional[NodeDefinition], default: None) – Definition of nodes in the graph.

• input_feature_names (Optional[List[str]], default: None) – Name of input feature columns.

• dtype (Optional[dtype], default: torch.float32) – data type for node features.

• perturbation_dict (Optional[Dict[str, float]], default: None) – Dictionary mapping a feature name to a standard deviation according to which the values for this feature should be randomly perturbed. Defaults to None.

• seed (Union[int, Generator, None], default: None) – seed or Generator used to randomly sample perturbations. Defaults to None.

• args (Any)

• kwargs (Any)

Return type:

object

class graphnet.models.data_representation.graphs.graphs.KNNGraphRRWP(*args, **kwargs)

Bases: GraphDefinition

KNN Graph with relative random walk probabilities (RRWP).

Identical to KNNGraph, but with five extra fields containing absolute and relative positional encoding using RRWP.

abs_pe = graph[“rrwp”] # RRWP absolute positional encoding values rrwp_val = graph[“rrwp_val”] # Non-zero values of the RRWP tensor rrwp_index = graph[“rrwp_index] # Corresponding row, col indices degree = graph[“deg”] # Degree of each node (num. of incoming edges)

Construct k-nn graph representation.

Parameters:

• detector (Detector) – Detector that represents your data.

• node_definition (Optional[NodeDefinition], default: None) – Definition of nodes in the graph.

• edge_definition (Optional[EdgeDefinition], default: None) – Definition of edges in the graph.

• input_feature_names (Optional[List[str]], default: None) – Name of input feature columns.

• dtype (Optional[dtype], default: torch.float32) – data type for node features.

• perturbation_dict (Optional[Dict[str, float]], default: None) – Dictionary mapping a feature name to a standard deviation according to which the values for this feature should be randomly perturbed. Defaults to None.

• seed (Union[int, Generator, None], default: None) – seed or Generator used to randomly sample perturbations. Defaults to None.

• nb_nearest_neighbours (int, default: 8) – Number of edges for each node. Defaults to 8.

• columns (List[int], default: [0, 1, 2]) – node feature columns used for distance calculation. Defaults to [0, 1, 2].

• walk_length (int, default: 8) – number of steps for the random walk. Defaults to 8.

• args (Any)

• kwargs (Any)

Return type:

object

forward(input_features, input_feature_names, **kwargs)

Forward pass.

Return type:

Data

Parameters:

• input_features (ndarray)

• input_feature_names (List[str])

class graphnet.models.data_representation.graphs.graphs.KNNGraphRWSE(*args, **kwargs)

Bases: GraphDefinition

KNN Graph with random walk structural encoding (RWSE).

Identical to KNNGraph but with an additional field containing the values obtained from RWSE. The encoding can be accessed via

rwse = graph[“rwse”] # random walk structural encoding

Construct k-nn graph representation.

Parameters:

• detector (Detector) – Detector that represents your data.

• node_definition (Optional[NodeDefinition], default: None) – Definition of nodes in the graph.

• edge_definition (Optional[EdgeDefinition], default: None) – Definition of edges in the graph.

• input_feature_names (Optional[List[str]], default: None) – Name of input feature columns.

• dtype (Optional[dtype], default: torch.float32) – data type for node features.

• perturbation_dict (Optional[Dict[str, float]], default: None) – Dictionary mapping a feature name to a standard deviation according to which the values for this feature should be randomly perturbed. Defaults to None.

• seed (Union[int, Generator, None], default: None) – seed or Generator used to randomly sample perturbations. Defaults to None.

• nb_nearest_neighbours (int, default: 8) – Number of edges for each node. Defaults to 8.

• columns (List[int], default: [0, 1, 2]) – node feature columns used for distance calculation. Defaults to [0, 1, 2].

• walk_length (int, default: 8) – number of steps for the random walk. Defaults to 8.

• args (Any)

• kwargs (Any)

Return type:

object

forward(input_features, input_feature_names, **kwargs)

Forward pass.

Return type:

Data

Parameters:

• input_features (ndarray)

• input_feature_names (List[str])