You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Describe the bug
A differential private reach value is calculated by
reach_dp = (reach + noise) / sampling_rate
Since the noise used in the system is discrete Laplace and so is integer, the differential private reach values outputted from the measurement system should always be multiple of the reciprocal of the sampling rate. However, based on the reach values gotten from the measurement system, it's not for both direct measurements and multi-publisher measurements.
Steps to reproduce
Use the MeasurementSystem tool to submit create measurement request for reach with either single publisher or multi-publisher and argument --vid-sampling-width == 0.01. Ex:
Additional context
We can confirm the direct measurement is not computed correctly as shown in EdpSimulator.kt#L835. Further investigation is needed for the multi-publisher case.
The text was updated successfully, but these errors were encountered:
Describe the bug
A differential private reach value is calculated by
Since the noise used in the system is discrete Laplace and so is integer, the differential private reach values outputted from the measurement system should always be multiple of the reciprocal of the sampling rate. However, based on the reach values gotten from the measurement system, it's not for both direct measurements and multi-publisher measurements.
Steps to reproduce
--vid-sampling-width == 0.01
. Ex:Component(s) affected
Kingdom, testing simulators
Version
6184fb9
Environment
QA env
Additional context
We can confirm the direct measurement is not computed correctly as shown in EdpSimulator.kt#L835. Further investigation is needed for the multi-publisher case.
The text was updated successfully, but these errors were encountered: