nexuml_library.scenarios.asd.conv_ae_toycaremu¶
Convolutional autoencoder ASD scenario on the ToyCarEmu DCASE 2026 machine.
conv_ae_toycaremu(
n_mels: int = 128,
hop_length: int = 512,
clip_num_samples: int = 160000,
latent_dim: int = 64,
channel_schedule: list[int] | None = None,
lr: float = 0.001,
batch_size: int = 32,
max_epochs: int = 25,
) -> ScenarioSpec
Convolutional AE anomaly detector trained on DCASE 2026 ToyCarEmu.
This is a narrow, single-machine didactic preset with an explicit decision rule in the loss stage.
Returns:
| Name | Type | Description |
|---|---|---|
ScenarioSpec |
ScenarioSpec
|
Assembled scenario with pipeline, training, data and evaluation. |