graph_datasets.datasets package

Submodules

graph_datasets.datasets.cola module

Datasets from the paper CoLA.

graph_datasets.datasets.cola.load_cola_data(dataset_name: str, directory: str = './data', verbosity: int = 0) → Tuple[DGLGraph, Tensor, int]

Load CoLA graphs.

Parameters:

• dataset_name (str) – Dataset name.

• directory (str, optional) – Raw dir for loading or saving. Defaults to DEFAULT_DATA_DIR=os.path.abspath(“./data”).

• verbosity (int, optional) – Output debug information. The greater, the more detailed. Defaults to 0.

Returns:

[graph, label, n_clusters]

Return type:

Tuple[dgl.DGLGraph, torch.Tensor, int]

graph_datasets.datasets.critical module

Datasets from the paper A Critical Look at the Evaluation of GNNs Under Heterophily: Are We Really Making Progress?.

graph_datasets.datasets.critical.load_critical_dataset(dataset_name: str, directory: str = './data', verbosity: int = 0) → Tuple[DGLGraph, Tensor, int]

Load graphs from A Critical Look at the Evaluation of GNNs Under Heterophily: Are We Really Making Progress?

Parameters:

• dataset_name (str) – Dataset name.

• directory (str, optional) – Raw dir for loading or saving. Defaults to DEFAULT_DATA_DIR=os.path.abspath(“./data”).

• verbosity (int, optional) – Output debug information. The greater, the more detailed. Defaults to 0.

Returns:

[graph, label, n_clusters]

Return type:

Tuple[dgl.DGLGraph, torch.Tensor, int]

graph_datasets.datasets.dgl module

Datasets from DGL.

graph_datasets.datasets.dgl.load_dgl_data(dataset_name: str, directory: str = './data', verbosity: int = 0) → Tuple[DGLGraph, Tensor, int]

Load DGL graphs.

Parameters:

• dataset_name (str) – Dataset name.

• directory (str, optional) – Raw dir for loading or saving. Defaults to DEFAULT_DATA_DIR=os.path.abspath(“./data”).

• verbosity (int, optional) – Output debug information. The greater, the more detailed. Defaults to 0.

Raises:

NotImplementedError – Dataset unknown.

Note

Chameleon, Squirrel, Actor, Cornell, Texas and Wisconsin are from Geom-GCN, which may be slightly different from heterophilous settings.

Returns:

[graph, label, n_clusters]

Return type:

Tuple[dgl.DGLGraph, torch.Tensor, int]

graph_datasets.datasets.linkx module

Datasets from the paper LINKX.

graph_datasets.datasets.linkx.even_quantile_labels(vals, n_classes, verbosity: int = 0)

partitions vals into n_classes by a quantile based split, where the first class is less than the 1/n_classes quantile, second class is less than the 2/n_classes quantile, and so on

vals is np array returns an np array of int class labels

graph_datasets.datasets.linkx.load_linkx_data(dataset_name: str, directory: str = './data', verbosity: int = 0) → Tuple[DGLGraph, Tensor, int]

Load LINKX graphs.

Parameters:

• dataset_name (str) – Dataset name.

• directory (str, optional) – Raw dir for loading or saving. Defaults to DEFAULT_DATA_DIR=os.path.abspath(“./data”).

• verbosity (int, optional) – Output debug information. The greater, the more detailed. Defaults to 0.

Returns:

[graph, label, n_clusters]

Return type:

Tuple[dgl.DGLGraph, torch.Tensor, int]

graph_datasets.datasets.linkx.load_linkx_github(dataset_name: str, directory: str = './data') → Tuple[DGLGraph, Tensor, int]

Load LINKX graphs.

Parameters:

• dataset_name (str) – Dataset name.

• directory (str, optional) – Raw dir for loading or saving. Defaults to DEFAULT_DATA_DIR=os.path.abspath(“./data”).

Returns:

[graph, label, n_clusters]

Return type:

Tuple[dgl.DGLGraph, torch.Tensor, int]

graph_datasets.datasets.linkx.load_wiki_data(directory: str = './data') → Tuple[DGLGraph, Tensor, int]

graph_datasets.datasets.linkx.load_twitch_gamers_data(task='mature', directory: str = './data', normalize: bool = True) → Tuple[DGLGraph, Tensor, int]

graph_datasets.datasets.linkx.load_fb100_data(dataset_name: str, directory: str = './data') → Tuple[DGLGraph, Tensor, int]

graph_datasets.datasets.linkx.load_arxiv_year_data(n_classes: int = 5, directory: str = './data', verbosity: int = 0) → Tuple[DGLGraph, Tensor, int]

graph_datasets.datasets.ogb module

Datasets from OGB.

graph_datasets.datasets.ogb.load_ogb_data(dataset_name: str, directory: str = './data', verbosity: int = 0) → Tuple[DGLGraph, Tensor, int]

Load OGB graphs.

Parameters:

• dataset_name (str) – Dataset name.

• directory (str, optional) – Raw dir for loading or saving. Defaults to DEFAULT_DATA_DIR=os.path.abspath(“./data”).

• verbosity (int, optional) – Output debug information. The greater, the more detailed. Defaults to 0.

Raises:

NotImplementedError – Dataset unknown.

Returns:

[graph, label, n_clusters]

Return type:

Tuple[dgl.DGLGraph, torch.Tensor, int]

graph_datasets.datasets.pyg module

Datasets from PyG.

graph_datasets.datasets.pyg.load_pyg_data(dataset_name: str, directory: str = './data', verbosity: int = 0) → Tuple[DGLGraph, Tensor, int]

Load pyG graphs.

Parameters:

• dataset_name (str) – Dataset name.

• directory (str, optional) – Raw dir for loading or saving. Defaults to DEFAULT_DATA_DIR=os.path.abspath(“./data”).

• verbosity (int, optional) – Output debug information. The greater, the more detailed. Defaults to 0.

Raises:

NotImplementedError – Dataset unknown.

Returns:

[graph, label, n_clusters]

Return type:

Tuple[dgl.DGLGraph, torch.Tensor, int]

Note

No row-normalization conducted.

graph_datasets.datasets.sdcn module

Datasets from the paper SDCN.

graph_datasets.datasets.sdcn.load_sdcn_data(dataset_name: str, directory: str = './data', verbosity: int = 0) → Tuple[DGLGraph, Tensor, int]

Load SDCN graphs.

Parameters:

• dataset_name (str) – Dataset name.

• directory (str, optional) – Raw dir for loading or saving. Defaults to DEFAULT_DATA_DIR=os.path.abspath(“./data”).

• verbosity (int, optional) – Output debug information. The greater, the more detailed. Defaults to 0.

Note

The last node of DBLP is an isolated node.

Returns:

[graph, label, n_clusters]

Return type:

Tuple[dgl.DGLGraph, torch.Tensor, int]

Module contents

Datasets from different sources