CLUSTERDatasetο
- class dgl.data.CLUSTERDataset(mode='train', raw_dir=None, force_reload=False, verbose=False, transform=None)[source]ο
Bases:
DGLBuiltinDatasetCLUSTER dataset for semi-supervised clustering task.
Each graph contains 6 SBM clusters with sizes randomly selected between [5, 35] and probabilities p = 0.55, q = 0.25. The graphs are of sizes 40 -190 nodes. Each node can take an input feature value in {0, 1, 2, β¦, 6} and values 1~6 correspond to classes 0~5 respectively, while value 0 means that the class of the node is unknown. There is only one labeled node that is randomly assigned to each community and most node features are set to 0.
Reference https://arxiv.org/pdf/2003.00982.pdf
Statistics:
Train examples: 10,000
Valid examples: 1,000
Test examples: 1,000
Number of classes for each node: 6
- Parameters:
mode (str) β Must be one of (βtrainβ, βvalidβ, βtestβ). Default: βtrainβ
raw_dir (str) β Raw file directory to download/contains the input data directory. Default: ~/.dgl/
force_reload (bool) β Whether to reload the dataset. Default: False
verbose (bool) β Whether to print out progress information. Default: False
transform (callable, optional) β A transform that takes in a
DGLGraphobject and returns a transformed version. TheDGLGraphobject will be transformed before every access.
Examples
>>> from dgl.data import CLUSTERDataset >>> >>> trainset = CLUSTERDataset(mode='train') >>> >>> trainset.num_classes 6 >>> len(trainset) 10000 >>> trainset[0] Graph(num_nodes=117, num_edges=4104, ndata_schemes={'label': Scheme(shape=(), dtype=torch.int16), 'feat': Scheme(shape=(), dtype=torch.int64)} edata_schemes={'feat': Scheme(shape=(1,), dtype=torch.float32)})