opendataval.dataval.csshap package

Submodules

opendataval.dataval.csshap.csshap module

class opendataval.dataval.csshap.csshap.ClassWiseShapley(*args, **kwargs)

Bases: DataEvaluator, ModelMixin

Class-wise shapley data valuation implementation

NOTE only categorical labels is a valid input to Class-Wise Shapley.

References

[1]

S. Schoch, H. Xu, and Y. Ji, CS-Shapley: Class-wise Shapley Values for Data Valuation in Classification arXiv.org, 2022. https://arxiv.org/abs/2211.06800.

Parameters

samplerSampler, optional

Sampler used to compute the marginal contributions. NOTE the sampler may not use a cache and cache_name should explicitly be passes None. Can be found in sampler, by default uses *args, **kwargs for TMCSampler but removes cache.

evaluate_data_values() → ndarray

Returns data values for CS-Shapley

input_data(x_train: Tensor, y_train: Tensor, x_valid: Tensor, y_valid: Tensor)

Store and transform input data for CS-Shapley.

Parameters

x_traintorch.Tensor

Data covariates

y_traintorch.Tensor

Data labels

x_validtorch.Tensor

Test+Held-out covariates

y_validtorch.Tensor

Test+Held-out labels

train_data_values(*args, **kwargs)

Uses sampler to trains model to find marginal contribs and data values.

For each class, we separate the training and validation data into in-class and out-class. Then we will compute the class-wise shapley values using the sampler. Finally, we record the shapley value in self.data_values.

Module contents