README.md

In schwenzfeier/udp: Statistically Useful Differential Privacy

UnbiasedPrivacy

Package that takes (almost) any function and returns a quantity of interest that is differentially private.

functions.R functions required to run the algorithm.

utils.R helper functions that calculate various quantities of interest or help with simulations.

Meg's To-Do List

• Clean up code: some legacy code has inconsistent variable names across functions, we will fix this

Items to note

• The random number generation is not currently differentially private
• The branching is a bit hacky right now
• We don't currently have robust error handling because we were testing a bunch of edge cases
• The one-sided vs. two-sided solver runs into small float issues that we should fix
• There is currently no testing for the functions
• We would like to add functionality so users can create their own functions and have some rules that must be followed to make sure the functions are allowable

schwenzfeier/udp documentation built on April 1, 2021, 3:39 a.m.