Learning junta distributions is computationally equivalent to LPN, and a new algorithm achieves near-optimal sample complexity O((k/epsilon^2)(2^k + log n)).
Learning and testing junta distributions with sub cube conditioning
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
New Statistical and Computational Results for Learning Junta Distributions
Learning junta distributions is computationally equivalent to LPN, and a new algorithm achieves near-optimal sample complexity O((k/epsilon^2)(2^k + log n)).