pydgc.pipelines

pydgc.pipelines.agcdrr_pipeline

class AGCDRRPipeline(args)[source]

Bases: BasePipeline

AGCDRR pipeline.

Parameters:

args (Namespace) – Arguments.

augment_data()[source]

Data augmentation.

Parameters:

self.data (Data) – PyG data object.

build_model()[source]

Build model.

Parameters:

self.cfg (CN) – Config object.

Returns:

Model object.

Return type:

DGCModel

pydgc.pipelines.base_pipeline

class BasePipeline(args)[source]

Bases: ABC

Standardized pipeline for deep graph clustering.

Parameters:

args (Namespace) – Arguments for setting values frequently changed.

load_config()[source]

load config from yaml

Parameters:

  • self.cfg_file_path (str) – Path to the config file.

  • self.dataset_name (str) – Name of the dataset.

load_logger()[source]

Load logger.

Parameters:

self.cfg (CN) – Config object.

load_dataset()[source]

Load dataset.

Parameters:

  • self.cfg (CN) – Config object.

  • self.dataset_name (str) – Name of the dataset.

abstract augment_data()[source]
abstract build_model()[source]

Build model.

Parameters:

self.cfg (CN) – Config object.

Returns:

Model object.

Return type:

DGCModel

evaluate(results)[source]

Evaluate model.

Parameters:

  • self.cfg (CN) – Config object.

  • results (dict) – Evaluation results.

visualize()[source]

Visualize results.

Parameters:

self.cfg (CN) – Config object.

run(pretrain=False, flag='TRAIN')[source]

Run pipeline.

Parameters:

  • self.cfg_file_path (str) – Path to the config file.

  • self.dataset_name (str) – Name of the dataset.

  • self.args (Namespace) – Arguments.

  • pretrain (bool) – Whether to pretrain the model.

  • flag (str) – Flag for logging.

pydgc.pipelines.ccgc_pipeline

preprocess_graph(adj, layer, norm='sym', renorm=True)[source]

Preprocess graph.

Parameters:

  • adj (sp.csr_matrix) – Adjacency matrix.

  • layer (int) – Number of layers.

  • norm (str) – Normalization method.

  • renorm (bool) – Whether to renormalize the adjacency matrix.

Returns:

List of preprocessed adjacency matrices.

Return type:

list

class CCGCPipeline(args)[source]

Bases: BasePipeline

CCGC pipeline.

Parameters:

args (Namespace) – Arguments.

augment_data()[source]
build_model()[source]

Build model.

Parameters:

self.cfg (CN) – Config object.

Returns:

Model object.

Return type:

DGCModel

pydgc.pipelines.daegc_pipeline

class DAEGCPipeline(args)[source]

Bases: BasePipeline

DAEGC pipeline.

Parameters:

args (Namespace) – Arguments.

augment_data()[source]
build_model()[source]

Build model.

Parameters:

self.cfg (CN) – Config object.

Returns:

Model object.

Return type:

DGCModel

pydgc.pipelines.dcrn_pipeline

normalize_adj(adj, self_loop=True, symmetry=False)[source]

Normalize the adj matrix.

Parameters:

  • adj (np.ndarray) – Input adj matrix.

  • self_loop (bool) – If add the self loop or not.

  • symmetry (bool) – Symmetry normalize or not.

Returns:

The normalized adj matrix.

Return type:

np.ndarray

diffusion_adj(adj, transport_rate=0.2)[source]

Graph diffusion.

Parameters:

  • adj (np.ndarray) – Input adj matrix.

  • transport_rate (float) – The transport rate.

Returns:

The graph diffusion.

Return type:

np.ndarray

sparse_mx_to_torch_sparse_tensor(sparse_mx)[source]

Convert a scipy sparse matrix to a torch sparse tensor.

Parameters:

sparse_mx (scipy.sparse) – Input sparse matrix.

Returns:

The torch sparse tensor.

Return type:

torch.sparse_coo_tensor

class DCRNPipeline(args)[source]

Bases: BasePipeline

DCRN pipeline.

Parameters:

args (Namespace) – Arguments.

augment_data()[source]
build_model()[source]

Build model.

Parameters:

self.cfg (CN) – Config object.

Returns:

Model object.

Return type:

DGCModel

pydgc.pipelines.dfcn_pipeline

normalize(mx)[source]

Row-normalize sparse matrix.

Parameters:

mx (scipy.sparse) – Input sparse matrix.

Returns:

Row-normalized sparse matrix.

Return type:

scipy.sparse

sparse_mx_to_torch_sparse_tensor(sparse_mx)[source]

Convert a scipy sparse matrix to a torch sparse tensor.

Parameters:

sparse_mx (scipy.sparse) – Input sparse matrix.

Returns:

The torch sparse tensor.

Return type:

torch.sparse_coo_tensor

class DFCNPipeline(args)[source]

Bases: BasePipeline

DFCN pipeline.

Parameters:

args (Namespace) – Arguments.

augment_data()[source]
build_model()[source]

Build model.

Parameters:

self.cfg (CN) – Config object.

Returns:

Model object.

Return type:

DGCModel

pydgc.pipelines.dgcluster_pipeline

class DGCLUSTERPipeline(args)[source]

Bases: BasePipeline

DGCLUSTER pipeline.

Parameters:

args (Namespace) – Arguments.

augment_data()[source]
build_model()[source]

Build model.

Parameters:

self.cfg (CN) – Config object.

Returns:

Model object.

Return type:

DGCModel

pydgc.pipelines.gae_pipeline

class GAEPipeline(args)[source]

Bases: BasePipeline

GAE pipeline.

Parameters:

args (Namespace) – Arguments.

augment_data()[source]
build_model()[source]

Build model.

Parameters:

self.cfg (CN) – Config object.

Returns:

Model object.

Return type:

DGCModel

pydgc.pipelines.gae_ssc_pipeline

class GAESSCPipeline(args)[source]

Bases: BasePipeline

GAE-SSC pipeline.

Parameters:

args (Namespace) – Arguments.

augment_data()[source]

Data augmentation

build_model()[source]

Build model.

Parameters:

self.cfg (CN) – Config object.

Returns:

Model object.

Return type:

DGCModel

pydgc.pipelines.hsan_pipeline

normalize_adj(adj, self_loop=True, symmetry=False)[source]

normalize the adj matrix :param adj: input adj matrix :param self_loop: if add the self loop or not :param symmetry: symmetry normalize or not :return: the normalized adj matrix

laplacian_filtering(A, X, t)[source]
class HSANPipeline(args)[source]

Bases: BasePipeline

HSAN pipeline.

Parameters:

args (Namespace) – Arguments.

augment_data()[source]

Data augmentation

build_model()[source]

Build model.

Parameters:

self.cfg (CN) – Config object.

Returns:

Model object.

Return type:

DGCModel

pydgc.pipelines.magi_batch_pipeline

get_sim(batch, adj, wt=20, wl=3)[source]

Get similarity matrix.

Parameters:

  • batch (torch.Tensor) – Batch indices.

  • adj (SparseTensor) – Adjacency matrix.

  • wt (int, optional) – Number of random walks. Defaults to 20.

  • wl (int, optional) – Length of random walks. Defaults to 3.

Returns:

Similarity matrix.

Return type:

torch.Tensor

get_mask(adj)[source]

Get mask matrix.

Parameters:

adj (SparseTensor) – Adjacency matrix.

Returns:

Mask matrix.

Return type:

SparseTensor

class MAGIBatchPipeline(args)[source]

Bases: BasePipeline

MAGI-Batch pipeline.

Parameters:

args (Namespace) – Arguments.

augment_data()[source]
build_model()[source]

Build model.

Parameters:

self.cfg (CN) – Config object.

Returns:

Model object.

Return type:

DGCModel

pydgc.pipelines.magi_pipeline

get_sim(batch, adj, wt=20, wl=3)[source]

Get similarity matrix.

Parameters:

  • batch (torch.Tensor) – Batch indices.

  • adj (SparseTensor) – Adjacency matrix.

  • wt (int, optional) – Number of random walks. Defaults to 20.

  • wl (int, optional) – Length of random walks. Defaults to 3.

Returns:

Similarity matrix.

Return type:

torch.Tensor

get_mask(adj)[source]

Get mask matrix.

Parameters:

adj (SparseTensor) – Adjacency matrix.

Returns:

Mask matrix.

Return type:

SparseTensor

class MAGIPipeline(args)[source]

Bases: BasePipeline

MAGI pipeline.

Parameters:

args (Namespace) – Arguments.

augment_data()[source]
build_model()[source]

Build model.

Parameters:

self.cfg (CN) – Config object.

Returns:

Model object.

Return type:

DGCModel

pydgc.pipelines.ns4gc_pipeline

class NS4GCPipeline(args)[source]

Bases: BasePipeline

NS4GC pipeline.

Parameters:

args (Namespace) – Arguments.

augment_data()[source]
build_model()[source]

Build model.

Parameters:

self.cfg (CN) – Config object.

Returns:

Model object.

Return type:

DGCModel

pydgc.pipelines.sdcn_pipeline

class SDCNPipeline(args)[source]

Bases: BasePipeline

SDCN pipeline.

Parameters:

args (Namespace) – Arguments.

augment_data()[source]
build_model()[source]

Build model.

Parameters:

self.cfg (CN) – Config object.

Returns:

Model object.

Return type:

DGCModel