`nexuml_library.scenarios.model.linear_ae`¶

nexuml_library.scenarios.model.linear_ae ¶

Linear autoencoder model scenario fragments.

linear_ae_reconstruction ¶

linear_ae_reconstruction(
    input_dim: int = 128,
    hidden_dims: list[int] | None = None,
    latent_dim: int = 8,
    activation: str = "torch.nn.ReLU",
    feature_key: str = "features",
) -> PipelineSpec

Create a PipelineSpec for a linear autoencoder with reconstruction loss.

Returns:

Name	Type	Description
`PipelineSpec`	`PipelineSpec`	Pipeline with linear encoder, decoder and reconstruction loss layers.

linear_ae_anomaly_detection ¶

linear_ae_anomaly_detection(
    input_dim: int = 128,
    hidden_dims: list[int] | None = None,
    latent_dim: int = 8,
    activation: str = "torch.nn.ReLU",
    feature_key: str = "features",
    score_reduction: str = "mean",
) -> PipelineSpec

Linear autoencoder with reconstruction loss and an anomaly score.

Returns:

Name	Type	Description
`PipelineSpec`	`PipelineSpec`	Pipeline with linear encoder, decoder, reconstruction loss and anomaly score layers.

linear_ae_lmbe ¶

linear_ae_lmbe(
    sample_rate: int = 16000,
    n_mels: int = 128,
    n_fft: int = 1024,
    hop_length: int = 512,
    time_frames: int = 128,
    hidden_dims: list[int] | None = None,
    latent_dim: int = 8,
    activation: str = "torch.nn.ReLU",
    score_reduction: str = "mean",
) -> PipelineSpec

Waveform -> LMBE -> linear autoencoder with an anomaly score.

Returns:

Name	Type	Description
`PipelineSpec`	`PipelineSpec`	Full waveform-to-reconstruction pipeline with LMBE feature extraction and linear autoencoder.

linear_ae_multiclass ¶

linear_ae_multiclass(
    input_dim: int = 128,
    hidden_dims: list[int] | None = None,
    latent_dim: int = 8,
    activation: str = "torch.nn.ReLU",
) -> PipelineSpec

Create a PipelineSpec for a linear AE with reconstruction + classification.

Returns:

Name	Type	Description
`PipelineSpec`	`PipelineSpec`	Pipeline with encoder, decoder, classification head, reconstruction loss, classification loss and metrics layers.

linear_ae_multilabel ¶

linear_ae_multilabel(
    input_dim: int = 128,
    hidden_dims: list[int] | None = None,
    latent_dim: int = 8,
    activation: str = "torch.nn.ReLU",
) -> PipelineSpec

Create a PipelineSpec for a linear AE with reconstruction + multilabel classification.

Returns:

Name	Type	Description
`PipelineSpec`	`PipelineSpec`	Pipeline with encoder, decoder, multilabel head, reconstruction loss and multilabel loss layers.

linear_ae_regression ¶

linear_ae_regression(
    input_dim: int = 128,
    hidden_dims: list[int] | None = None,
    latent_dim: int = 8,
    num_outputs: int = 3,
    activation: str = "torch.nn.ReLU",
) -> PipelineSpec

Create a PipelineSpec for a linear AE with reconstruction + regression.

Returns:

Name	Type	Description
`PipelineSpec`	`PipelineSpec`	Pipeline with encoder, decoder, regression head, reconstruction loss and regression loss layers.

nexuml_library.scenarios.model.linear_ae¶