conditional

The Conditional Adaption approaches are derived from the [MMD] [ rul_adapt.approach.mmd] and DANN approaches. They apply their respective adaption loss not only to the whole data but also separately to subsets of the data with a ConditionalAdaptionLoss. Fuzzy sets with rectangular membership functions define these subsets.

Both variants were introduced by Cheng et al. in 2021.

ConditionalDannApproach

Bases: AdaptionApproach

The conditional DANN approach uses a marginal and several conditional domain discriminators. The features are produced by a shared feature extractor. The loss in the domain discriminators is binary cross-entropy.

The regressor and domain discriminators need the same number of input units as the feature extractor has output units. The discriminators are not allowed to have an activation function on their last layer and need to use only a single output neuron because BCEWithLogitsLoss is used.

Examples:

>>> from rul_adapt import model
>>> from rul_adapt import approach
>>> feat_ex = model.CnnExtractor(1, [16, 16, 1], 10, fc_units=16)
>>> reg = model.FullyConnectedHead(16, [1])
>>> disc = model.FullyConnectedHead(16, [8, 1], act_func_on_last_layer=False)
>>> cond_dann = approach.ConditionalDannApproach(1.0, 0.5, [(0, 1)])
>>> cond_dann.set_model(feat_ex, reg, disc)

__init__(dann_factor, dynamic_adaptive_factor, fuzzy_sets, loss_type='mae', rul_score_mode='phm08', evaluate_degraded_only=False, **optim_kwargs)

Create a new conditional DANN approach.

The strength of the domain discriminator's influence on the feature extractor is controlled by the dann_factor. The higher it is, the stronger the influence. The dynamic_adaptive_factor controls the balance between the marginal and conditional DANN loss.

The domain discriminator is set by the set_model function together with the feature extractor and regressor. For more information, see the approach module page.

For more information about the possible optimizer keyword arguments, see here.

Parameters:

Name	Type	Description	Default
dann_factor	float	Strength of the domain DANN loss.	required
dynamic_adaptive_factor	float	Balance between the marginal and conditional DANN loss.	required
fuzzy_sets	List[Tuple[float, float]]	Fuzzy sets for the conditional DANN loss.	required
loss_type	Literal['mse', 'rmse', 'mae']	The type of regression loss, either 'mse', 'rmse' or 'mae'.	'mae'
rul_score_mode	Literal['phm08', 'phm12']	The mode for the val and test RUL score, either 'phm08' or 'phm12'.	'phm08'
**optim_kwargs	Any	Keyword arguments for the optimizer, e.g. learning rate.	{}

configure_optimizers()

Configure an Adam optimizer.

forward(inputs)

Predict the RUL values for a batch of input features.

set_model(feature_extractor, regressor, domain_disc=None, *args, **kwargs)

Set the feature extractor, regressor, and domain discriminator for this approach.

The discriminator is not allowed to have an activation function on its last layer and needs to use only a single output neuron. It is wrapped by a DomainAdversarialLoss.

A copy of the discriminator is used for each conditional loss governing a fuzzy set.

Parameters:

Name	Type	Description	Default
feature_extractor	Module	The feature extraction network.	required
regressor	Module	The RUL regression network.	required
domain_disc	Optional[Module]	The domain discriminator network. Copied for each fuzzy set.	None

test_step(batch, batch_idx, dataloader_idx)

Execute one test step.

The batch argument is a list of two tensors representing features and labels. A RUL prediction is made from the features and the validation RMSE and RUL score are calculated. The metrics recorded for dataloader idx zero are assumed to be from the source domain and for dataloader idx one from the target domain. The metrics are written to the configured logger under the prefix test.

Parameters:

Name	Type	Description	Default
batch	List[Tensor]	A list containing a feature and a label tensor.	required
batch_idx	int	The index of the current batch.	required
dataloader_idx	int	The index of the current dataloader (0: source, 1: target).	required

training_step(batch, batch_idx)

Execute one training step.

The batch argument is a list of three tensors representing the source features, source labels and target features. Both types of features are fed to the feature extractor. Then the regression loss for the source domain, the MMD loss and the conditional MMD loss are computed. The regression, MMD, conditional MMD and combined loss are logged.

Parameters:

Name	Type	Description	Default
batch	List[Tensor]	A list of a source feature, source label and target feature tensors.	required
batch_idx	int	The index of the current batch.	required

Returns: The combined loss.

validation_step(batch, batch_idx, dataloader_idx)

Execute one validation step.

The batch argument is a list of two tensors representing features and labels. A RUL prediction is made from the features and the validation RMSE and RUL score are calculated. The metrics recorded for dataloader idx zero are assumed to be from the source domain and for dataloader idx one from the target domain. The metrics are written to the configured logger under the prefix val.

Parameters:

Name	Type	Description	Default
batch	List[Tensor]	A list containing a feature and a label tensor.	required
batch_idx	int	The index of the current batch.	required
dataloader_idx	int	The index of the current dataloader (0: source, 1: target).	required

ConditionalMmdApproach

Bases: AdaptionApproach

The conditional MMD uses a combination of a marginal and conditional MML loss to adapt a feature extractor to be used with the source regressor.

The regressor needs the same number of input units as the feature extractor has output units.

Examples:

>>> from rul_adapt import model
>>> from rul_adapt import approach
>>> feat_ex = model.CnnExtractor(1, [16, 16, 1], 10, fc_units=16)
>>> reg = model.FullyConnectedHead(16, [1])
>>> cond_mmd = approach.ConditionalMmdApproach(0.01, 5, 0.5, [(0, 1)])
>>> cond_mmd.set_model(feat_ex, reg)

__init__(mmd_factor, num_mmd_kernels, dynamic_adaptive_factor, fuzzy_sets, loss_type='mae', rul_score_mode='phm08', evaluate_degraded_only=False, **optim_kwargs)

Create a new conditional MMD approach.

The strength of the influence of the MMD loss on the feature extractor is controlled by the mmd_factor. The higher it is, the stronger the influence. The dynamic adaptive factor controls the balance between the marginal MMD and conditional MMD losses.

For more information about the possible optimizer keyword arguments, see here.

Parameters:

Name	Type	Description	Default
mmd_factor	float	The strength of the MMD loss' influence.	required
num_mmd_kernels	int	The number of kernels for the MMD loss.	required
dynamic_adaptive_factor	float	The balance between marginal and conditional MMD.	required
fuzzy_sets	List[Tuple[float, float]]	The fuzzy sets for the conditional MMD loss.	required
loss_type	Literal['mse', 'rmse', 'mae']	The type of regression loss, either 'mse', 'rmse' or 'mae'.	'mae'
rul_score_mode	Literal['phm08', 'phm12']	The mode for the val and test RUL score, either 'phm08' or 'phm12'.	'phm08'
evaluate_degraded_only	bool	Whether to only evaluate the RUL score on degraded samples.	False
**optim_kwargs	Any	Keyword arguments for the optimizer, e.g. learning rate.	{}

configure_optimizers()

Configure an Adam optimizer.

forward(inputs)

Predict the RUL values for a batch of input features.

test_step(batch, batch_idx, dataloader_idx)

Execute one test step.

The batch argument is a list of two tensors representing features and labels. A RUL prediction is made from the features and the validation RMSE and RUL score are calculated. The metrics recorded for dataloader idx zero are assumed to be from the source domain and for dataloader idx one from the target domain. The metrics are written to the configured logger under the prefix test.

Parameters:

Name	Type	Description	Default
batch	List[Tensor]	A list containing a feature and a label tensor.	required
batch_idx	int	The index of the current batch.	required
dataloader_idx	int	The index of the current dataloader (0: source, 1: target).	required

training_step(batch, batch_idx)

Execute one training step.

The batch argument is a list of three tensors representing the source features, source labels and target features. Both types of features are fed to the feature extractor. Then the regression loss for the source domain, the MMD loss and the conditional MMD loss are computed. The regression, MMD, conditional MMD and combined loss are logged.

Parameters:

Name	Type	Description	Default
batch	List[Tensor]	A list of a source feature, source label and target feature tensors.	required
batch_idx	int	The index of the current batch.	required

Returns: The combined loss.

validation_step(batch, batch_idx, dataloader_idx)

Execute one validation step.

The batch argument is a list of two tensors representing features and labels. A RUL prediction is made from the features and the validation RMSE and RUL score are calculated. The metrics recorded for dataloader idx zero are assumed to be from the source domain and for dataloader idx one from the target domain. The metrics are written to the configured logger under the prefix val.

Parameters:

Name	Type	Description	Default
batch	List[Tensor]	A list containing a feature and a label tensor.	required
batch_idx	int	The index of the current batch.	required
dataloader_idx	int	The index of the current dataloader (0: source, 1: target).	required