NeuralTSNE.TSNE.ParametricTSNE package

Submodules

NeuralTSNE.TSNE.ParametricTSNE.parametric_tsne module

class ParametricTSNE(loss_fn: str, perplexity: int, batch_size: int, early_exaggeration_epochs: int, early_exaggeration_value: float, max_iterations: int, n_components: int | None = None, features: int | None = None, multipliers: List[float] | None = None, n_jobs: int = 0, tolerance: float = 1e-05, force_cpu: bool = False, model: NeuralNetwork | Module | OrderedDict | None = None)[source]View on GitHub

Bases: object

Parametric t-SNE implementation using a neural network model.

Parameters:

  • loss_fn (str) – Loss function for t-SNE. Currently supports kl_divergence.

  • perplexity (int) – Perplexity parameter for t-SNE.

  • batch_size (int) – Batch size for training.

  • early_exaggeration_epochs (int) – Number of epochs for early exaggeration.

  • early_exaggeration_value (float) – Early exaggeration factor.

  • max_iterations (int) – Maximum number of iterations for optimization.

  • n_components (int, optional) – Number of components in the output. Defaults to None.

  • features (int, optional) – Number of input features. Defaults to None.

  • multipliers (List[float], optional) – List of multipliers for hidden layers in the neural network. Defaults to None.

  • n_jobs (int, optional) – Number of workers for data loading. Defaults to 0.

  • tolerance (float, optional) – Tolerance level for convergence. Defaults to 1e-5.

  • force_cpu (bool, optional) – Force using CPU even if GPU is available. Defaults to False.

  • model (Union[NeuralNetwork, nn.Module, OrderedDict], optional) – Predefined model. Defaults to None.

create_dataloaders(train: Dataset, test: Dataset) → Tuple[DataLoader | None, DataLoader | None][source]View on GitHub

Create dataloaders for training and testing sets.

Parameters:

  • train (Dataset) – Training dataset.

  • test (Dataset) – Testing dataset.

Returns:

Tuple containing training and testing dataloaders.

Return type:

Tuple[DataLoader | None, DataLoader | None]

read_model(filename: str)[source]View on GitHub

Load the model’s state dictionary from a file.

Parameters:

filename (str) – Name of the file to load the model.

save_model(filename: str)[source]View on GitHub

Save the model’s state dictionary to a file.

Parameters:

filename (str) – Name of the file to save the model.

set_loss_fn(loss_fn: str) → Callable[source]View on GitHub

Set the loss function based on the provided string.

Parameters:

loss_fn (str) – String indicating the desired loss function.

Returns:

Corresponding loss function.

Return type:

Callable

Note

Currently supports kl_divergence as the loss function.

split_dataset(X: Tensor, y: Tensor = None, train_size: float = None, test_size: float = None) → Tuple[DataLoader | None, DataLoader | None][source]View on GitHub

Split the dataset into training and testing set

Parameters:

  • X (torch.Tensor) – Input data tensor.

  • y (torch.Tensor, optional) – Target tensor. Default is None.

  • train_size (float, optional) – Proportion of the dataset to include in the training set.

  • test_size (float, optional) – Proportion of the dataset to include in the testing set.

Returns:

Tuple containing training and testing dataloaders.

Return type:

Tuple[DataLoader | None, DataLoader | None]

Note

Splits the input data into training and testing sets, and returns corresponding dataloaders.

Module contents