Autoencoder

The LitAutoEncoder is a PyTorch Lightning module designed for unsupervised learning tasks. It consists of an encoder and a decoder network. The autoencoder's role in the project is to learn a compressed, dense representation of the input data (encoding) and then reconstruct the input data from this representation (decoding). This process helps in understanding the underlying structure of the data and is useful for tasks like anomaly detection, data denoising, and dimensionality reduction.

Key Features

• Encoder: Compresses input data into a latent representation.
• Decoder: Reconstructs the input data from the latent representation.
• Loss Function: Mean Squared Error (MSE) is used to measure reconstruction quality.

Autoencoder Class API

uv_datascience_project_template.lit_auto_encoder

LitAutoEncoder(encoder, decoder)

Bases: LightningModule

A simple autoencoder model.

PARAMETER	DESCRIPTION
encoder	The encoder component, responsible for encoding input data. TYPE: Sequential
decoder	The decoder component, responsible for decoding encoded data. TYPE: Sequential

Source code in src/uv_datascience_project_template/lit_auto_encoder.py

def __init__(self, encoder: nn.Sequential, decoder: nn.Sequential) -> None:
    super().__init__()
    self.encoder = encoder
    self.decoder = decoder

configure_optimizers()

Configure the Adam optimizer.

Source code in src/uv_datascience_project_template/lit_auto_encoder.py

def configure_optimizers(self) -> optim.Adam:
    """Configure the Adam optimizer."""
    optimizer = optim.Adam(self.parameters(), lr=1e-3)
    return optimizer

training_step(batch, batch_idx)

Performs a single training step for the model.

PARAMETER	DESCRIPTION
batch	A tuple containing the input data (x) and the corresponding labels (y). TYPE: Tuple[Tensor, Tensor]
batch_idx	The index of the current batch. TYPE: int

RETURNS	DESCRIPTION
Tensor	The computed loss for the current training step. TYPE: Tensor

Source code in src/uv_datascience_project_template/lit_auto_encoder.py

def training_step(self, batch: Tuple[Tensor, Tensor], batch_idx: int) -> Tensor:
    """Performs a single training step for the model.

    Args:
        batch (Tuple[Tensor, Tensor]): A tuple containing the input data (x) and
            the corresponding labels (y).
        batch_idx (int): The index of the current batch.

    Returns:
        Tensor: The computed loss for the current training step.
    """

    x, y = batch
    x = x.view(x.size(0), -1)
    z = self.encoder(x)
    x_hat = self.decoder(z)
    loss = nn.functional.mse_loss(x_hat, x)
    # Logging to TensorBoard (if installed) by default
    # self.log("train_loss", loss)
    return loss