deepof.hypermodels.Contrastive

class deepof.hypermodels.Contrastive(input_shape: tuple, latent_dim: int, learn_rate: float = 0.001, edge_feature_shape: tuple | None = None, use_gnn: bool = False, adjacency_matrix: ndarray | None = None)

Hyperparameter tuning pipeline for deepof.models.Contrastive.

__init__(input_shape: tuple, latent_dim: int, learn_rate: float = 0.001, edge_feature_shape: tuple | None = None, use_gnn: bool = False, adjacency_matrix: ndarray | None = None)

Contrastive hypermodel for hyperparameter tuning.

Parameters:

• input_shape (tuple) – shape of the input tensor.

• latent_dim (int) – dimension of the latent space.

• learn_rate (float) – learning rate for the optimizer.

• edge_feature_shape (tuple) – shape of the edge feature tensor.

• use_gnn (bool) – whether to use a graph neural network to encode the input data.

• adjacency_matrix (np.ndarray) – adjacency matrix of the graph.

Methods

__init__(input_shape, latent_dim[, ...])	Contrastive hypermodel for hyperparameter tuning.
build(hp)	Override Hypermodel's build method.
declare_hyperparameters(hp)
fit(hp, model, *args, **kwargs)	Train the model.
get_hparams(hp)	Retrieve hyperparameters to tune, including the encoder type and the weight of the kmeans loss.

__init__(input_shape: tuple, latent_dim: int, learn_rate: float = 0.001, edge_feature_shape: tuple | None = None, use_gnn: bool = False, adjacency_matrix: ndarray | None = None)

Contrastive hypermodel for hyperparameter tuning.

Parameters:

• input_shape (tuple) – shape of the input tensor.

• latent_dim (int) – dimension of the latent space.

• learn_rate (float) – learning rate for the optimizer.

• edge_feature_shape (tuple) – shape of the edge feature tensor.

• use_gnn (bool) – whether to use a graph neural network to encode the input data.

• adjacency_matrix (np.ndarray) – adjacency matrix of the graph.