egc.model.node_embedding.structure_learning package

Submodules

egc.model.node_embedding.structure_learning.sublime module

Towards Unsupervised Deep Graph Structure Learning https://shiruipan.github.io/publication/www-22-liu/www-22-liu.pdf

class egc.model.node_embedding.structure_learning.sublime.SUBLIME(nfeats, n_clusters, sparse: int = 0, type_learner: str = 'fgp', k: int = 20, sim_function: str = 'cosine', activation_learner: str = 'relu', nlayers: int = 2, hidden_dim: int = 512, rep_dim: int = 256, proj_dim: int = 256, dropout: float = 0.5, dropedge_rate: float = 0.5, lr: float = 0.001, w_decay: float = 0.0, epochs: int = 2500, maskfeat_rate_anchor: float = 0.8, maskfeat_rate_learner: float = 0.1, contrast_batch_size: int = 0, tau: float = 0.9999, c: int = 0, eval_freq: int = 100, n_clu_trials: int = 10)

Bases: Module

SUBLIME:Towards Unsupervised Deep Graph Structure Learning

Args:

forward(features, anchor_adj)

Forward Propagation

fit(adj_csr, features)

Fitting

Parameters:

• adj_csr (sp.lil_matrix) – adj sparse matrix.

• features (torch.Tensor) – features.

training: bool

get_embedding()

Get the embeddings.

get_memberships()

Get memberships

Returns:

memberships

Return type:

np.ndarray

Module contents

Graph Structure Learning Methods