PipelineDP is a Python framework for applying differentially private aggregations to large datasets using batch processing systems such as Apache Spark, Apache Beam, and more.
This PR implements MeanMechanism class (in dp_computations.py). This class is used for computing DP Mean in MeanCombiner. This is the part of the refactoring for making adding noise in more OOP style. This refactoring is needed for supporting PLD and l1 sensitivities in future.
Additional possibility which adds this refactoring it's possible to have more detailed description in explain computation reports about computin DP mean, here is an example:
_5. DP mean computation:
a. Computed 'normalized_sum' = sum of (value - 3.0)
b. Applied to 'count' Laplace mechanism: parameter=6.0 eps=0.3333333333333333 l1_sensitivity=2.0
c. Applied to 'normalized_sum' Laplace mechanism: parameter=12.0 eps=0.3333333333333333 l1sensitivity=4.0
This PR implements
MeanMechanismclass (in dp_computations.py). This class is used for computing DP Mean inMeanCombiner. This is the part of the refactoring for making adding noise in more OOP style. This refactoring is needed for supporting PLD and l1 sensitivities in future.Additional possibility which adds this refactoring it's possible to have more detailed description in explain computation reports about computin DP mean, here is an example:
_5. DP mean computation: a. Computed 'normalized_sum' = sum of (value - 3.0) b. Applied to 'count' Laplace mechanism: parameter=6.0 eps=0.3333333333333333 l1_sensitivity=2.0 c. Applied to 'normalized_sum' Laplace mechanism: parameter=12.0 eps=0.3333333333333333 l1sensitivity=4.0