Update StandardizedHistogramEstimator to (i) include the delta parameter (ii) support privacy budget tracking (iii) add more tests.
Will add more tests once Pasin updates Gaussian and DiscreteGaussian noisers.
@matthewclegg I'm trying to support privacy budget tracking here, but not sure if my code design is reasonable.
@pasin30055 FYI of what we're doing. No need to carefully review it at this moment. An open question here is: how to composite the privacy budget for both reach and frequency estimation? You have result for discrete Laplacian noises in our paper, but how about Gaussian noises? We can discuss more in Tuesday's meeting.
Update StandardizedHistogramEstimator to (i) include the delta parameter (ii) support privacy budget tracking (iii) add more tests.
Will add more tests once Pasin updates Gaussian and DiscreteGaussian noisers.
@matthewclegg I'm trying to support privacy budget tracking here, but not sure if my code design is reasonable.
@pasin30055 FYI of what we're doing. No need to carefully review it at this moment. An open question here is: how to composite the privacy budget for both reach and frequency estimation? You have result for discrete Laplacian noises in our paper, but how about Gaussian noises? We can discuss more in Tuesday's meeting.