Hi team,
I am trying to find the implementation of algorithm mentioned section 5.1.1 Automatic bounds determination of paper in JAVA.
Lines from the paper
"This differs from standard workflows, and more importantly, it requires prior knowledge of the data that an analyst might not have. To remove this hurdle, we designed an aggregation function which can be sequentially composed with our previously introduced ε DP functions to automatically compute bounds that minimize accuracy loss. Call this function APPROX_BOUNDS(col)."
I am looking at the ApproximateBounds implementation but it requires a parameter maxContributionBound which analysts do not have any idea. Is there any other implementation of approximate bound wherein it doesn't require maxContributionBound value and implicitly determines it in a differentially private manner to find approximate bound for a column?
I might have asked similar question in past but once again reframing here for clarity.
Hi team, I am trying to find the implementation of algorithm mentioned section 5.1.1 Automatic bounds determination of paper in JAVA.
Lines from the paper "This differs from standard workflows, and more importantly, it requires prior knowledge of the data that an analyst might not have. To remove this hurdle, we designed an aggregation function which can be sequentially composed with our previously introduced ε DP functions to automatically compute bounds that minimize accuracy loss. Call this function APPROX_BOUNDS(col)."
I am looking at the ApproximateBounds implementation but it requires a parameter maxContributionBound which analysts do not have any idea. Is there any other implementation of approximate bound wherein it doesn't require maxContributionBound value and implicitly determines it in a differentially private manner to find approximate bound for a column?
I might have asked similar question in past but once again reframing here for clarity.