Paper gives near-optimal private learning of discrete distributions
Polynomial-time pure and approximate DP algorithms match known sample rates up to logarithmic factors for batched user data.
Stories per day · 30 days
1 stories, newest first · one card per story