A distribution class exists that is learnable non-privately in TV distance with finite samples but not under differential privacy, weakly refuting Ashtiani's conjecture.
Bounds on the sample complexity for private learning and private data release
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.DS 1years
2024 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Not All Learnable Distribution Classes are Privately Learnable
A distribution class exists that is learnable non-privately in TV distance with finite samples but not under differential privacy, weakly refuting Ashtiani's conjecture.