Score: 1

Learning CNF formulas from uniform random solutions in the local lemma regime

Published: November 4, 2025 | arXiv ID: 2511.02487v1

By: Weiming Feng , Xiongxin Yang , Yixiao Yu and more

Potential Business Impact:

Teaches computers to understand complex logic puzzles.

Business Areas:
A/B Testing Data and Analytics

We study the problem of learning a $n$-variables $k$-CNF formula $\Phi$ from its i.i.d. uniform random solutions, which is equivalent to learning a Boolean Markov random field (MRF) with $k$-wise hard constraints. Revisiting Valiant's algorithm (Commun. ACM'84), we show that it can exactly learn (1) $k$-CNFs with bounded clause intersection size under Lov\'asz local lemma type conditions, from $O(\log n)$ samples; and (2) random $k$-CNFs near the satisfiability threshold, from $\widetilde{O}(n^{\exp(-\sqrt{k})})$ samples. These results significantly improve the previous $O(n^k)$ sample complexity. We further establish new information-theoretic lower bounds on sample complexity for both exact and approximate learning from i.i.d. uniform random solutions.

Country of Origin
πŸ‡­πŸ‡° πŸ‡ΊπŸ‡Έ πŸ‡¨πŸ‡³ Hong Kong, China, United States

Page Count
64 pages

Category
Computer Science:
Data Structures and Algorithms