class ModelTrainEmbeddings(data: Graph, loss_function: Dict, device: torch.device, conv: str = 'GCN', tune_out: bool = False)

Bases: object

Model for training Net, which building embeddings for Geom-GCN layer

Parameters:

• data – (Graph): Input Graph

• loss_function – (dict): Dict of parameters of unsupervised loss function

• conv – (str): Name of convolution (default:’GCN’)

• device – (device): Either ‘cuda’ or ‘cpu’ (default:’cuda’)

• tune_out – (bool): Flag if you want tuning out layer or if it 2 for GeomGCN

run(params: Dict) → torch_geometric.typing.Tensor

Learn embeddings

Parameters:

params – dict[str,float,int,float]: Parameters for learning: size of hidden layer, dropout, number of layers for the model, learning rate

Returns:

(Tensor): The output embeddings

class OptunaTrainEmbeddings(data: Graph, loss_function: Dict, device: torch.device, conv: str = 'GCN', tune_out: bool = False)

Bases: ModelTrainEmbeddings

Model for training Net, wcich building embeddings for Geom-GCN layer

Parameters:

• loss_function – (dict): Dict of parameters of unsupervised loss function

• conv – (str): Name of convolution (default:’GCN’)

• device – (device): Either ‘cuda’ or ‘cpu’ (default:’cuda’)

run(number_of_trials: int) → Dict[Any, Any]

Tuning parameters for learning embeddings

Parameters:

number_of_trials – (int): Number of trials for optuna

Returns:

(dict[str,float,int,float]): Learned parameters: size of hidden layer, dropout, number of layers for the model, learning rate