functional

get_latent_align(dataset, source_fd, target_fd, xjtu_sy_subtask=None, **trainer_kwargs)

Construct a Latent Alignment approach for the selected dataset with the original hyperparameters.

For the XJTU-SY task only FD001 and FD002 are available. The subtask controls if the bearing with the id 1 or 2 is used as the target data.

Examples:

>>> import rul_adapt
>>> dm, latent, trainer = rul_adapt.construct.get_latent_align("cmapss", 3, 1)
>>> trainer.fit(latent, dm)
>>> trainer.test(latent, dm)

Parameters:

Name	Type	Description	Default
dataset	Literal['cmapss', 'xjtu-sy']	The dataset to use.	required
source_fd	int	The source FD.	required
target_fd	int	The target FD.	required
xjtu_sy_subtask	Optional[int]	The subtask for the XJTU-SY (either 1 or 2).	None
trainer_kwargs	Any	Overrides for the trainer class.	{}

Returns: dm: The data module for adaption of the sub-datasets. dann: The Latent Alignment approach with feature extractor and regressor. trainer: The trainer object.

get_latent_align_config(dataset, source_fd, target_fd, xjtu_sy_subtask=None)

Get a configuration for the Latent Alignment approach.

For the XJTU-SY task only FD001 and FD002 are available. The subtask controls if the bearing with the id 1 or 2 is used as the target data. The configuration can be modified and fed to latent_align_from_config to create the approach.

Parameters:

Name	Type	Description	Default
dataset	Literal['cmapss', 'xjtu-sy']	The dataset to use.	required
source_fd	int	The source FD.	required
target_fd	int	The target FD.	required
xjtu_sy_subtask	Optional[int]	The subtask for the XJTU-SY (either 1 or 2).	None

Returns: The Latent Alignment configuration.

latent_align_from_config(config, **trainer_kwargs)

Construct a Latent Alignment approach from a configuration.

The configuration can be created by calling get_latent_align_config.

Parameters:

Name	Type	Description	Default
config	DictConfig	The Latent Alignment configuration.	required
trainer_kwargs	Any	Overrides for the trainer class.	{}

Returns: dm: The data module for adaption of the sub-datasets. dann: The Latent Alignment approach with feature extractor, regressor. trainer: The trainer object.