Exponential Mechanism
Mechanism Design via Differential Privacy (2007)
Frank McSherry, Kunal Talwar
https://programming-dp.com/notebooks/ch9.html
有限セット内のそれぞれの要素に対するアウトプット効用をscoring functionとして定め、εで近似的にスコアが最大化するように要素をアウトプットする。
アウトプットは常に、有限セットの中に含まれる要素となる。（noiseを加える結果がmake senseしない場合に有効）
A Better Privacy Analysis of the Exponential Mechanism
algorithms for private selection
PRIVATE SELECTION FROM PRIVATE CANDIDATES (2018)
refs
http://www-infosec.ist.osaka-u.ac.jp/~yasunaga/infosec/resume20160705_full.pdf
https://www.cis.upenn.edu/~aaroth/courses/slides/Lecture3.pdf
implementaion notes
https://github.com/cilvento/b2dp
Implementing the Exponential Mechanism with Base-2 Differential Privacy
addressing the concerns associated with translating arbitrary- or infinite-precision theoretical mechanisms to the reality of floating point or fixed precision
Base Exponential (Candidates) -- Implementation · Issue #324 · opendp/opendp · GitHub
says Christina's base-2 exponential would be a more robust solution for noising, but I think that belongs in its own issue.
https://github.com/opendp/smartnoise-core/blob/develop/runtime-rust/src/utilities/mechanisms.rs#L210-L265
#PPDM