want a small programming exercise in privacy-preserving statistics? Show more
Say you have lots of data about people liking things, and you want to share "people who like A also like B" (the "2-way marginals") without breaching individuals' privacy. This paper gives one method:
Exercise: Read at least section 2, then implement step 2 of the algorithm in section 3.3. The notation assumes a strong math background but it's really just arrays/sets and maps/dicts. Ask me for details!