Testing Formula Satisfaction

We study the query complexity of testing for properties defined by read once formulas, as instances of {\em massively parametrized properties}, and prove several testability and non-testability results. First we prove the testability of any property accepted by a Boolean read-once formula involving any bounded arity gates, with a number of queries exponential in $\epsilon$, doubly exponential in the arity, and independent of all other parameters. When the gates are limited to being monotone, we prove that there is an {\em estimation} algorithm, that outputs an approximation of the distance of the input from satisfying the property. For formulas only involving And/Or gates, we provide a more efficient test whose query complexity is only quasipolynomial in $\epsilon$. On the other hand, we show that such testability results do not hold in general for formulas over non-Boolean alphabets; specifically we construct a property defined by a read-once arity $2$ (non-Boolean) formula over an alphabet of size $4$, such that any $1/4$-test for it requires a number of queries depending on the formula size. We also present such a formula over an alphabet of size $5$ that additionally satisfies a strong monotonicity condition.

💡 Research Summary

This paper conducts a comprehensive investigation into the query complexity of testing properties defined by read-once formulas, situated within the framework of massively parametrized property testing. The central question is: given a read-once formula Φ (known fully to the tester) and an assignment σ to its variables (accessible only via queries), how many queries are needed to distinguish between the case where σ satisfies Φ and the case where σ is ε-far from satisfying any such assignment?

The authors establish a detailed landscape of testability and non-testability. The first main result proves that for any read-once Boolean formula using gates of bounded arity k, the property SAT(Φ) is testable. The query complexity is exponential in a polynomial of 1/ε and doubly exponential in k, but crucially independent of the formula size or other structural parameters. This demonstrates a surprising uniformity: as long as the gate arity is bounded, even complex Boolean functions lead to testable properties.

The second result strengthens this for monotone gates. If all gates in Φ are monotone Boolean functions, there exists not just a tester, but an estimator that, with high probability, outputs an approximation of the actual distance of σ from SAT(Φ), within an additive error of ε. This highlights the power of monotonicity in enabling more refined property analysis.

The third result shows a significant efficiency improvement for a common special case. For formulas consisting only of AND and OR gates (of any arity), a much more efficient tester exists, with query complexity that is only quasipolynomial in 1/ε (specifically, of the form exp(O(log²(1/ε)))). This contrasts with the exponential dependence for general bounded-arity gates.

The fourth and perhaps most striking set of results reveals a fundamental limitation: the positive results do not extend to non-Boolean alphabets. The authors construct an explicit family of read-once formulas over an alphabet of size 4, using just one type of binary symmetric gate, for which any 1/4-test requires a number of queries that grows with the formula size (specifically, at least logarithmic in the depth for adaptive testers, and linear in depth for non-adaptive ones). This establishes that testability is not a general property of read-once formulas but is specific to the Boolean domain. They further strengthen this negative result by constructing a similar non-testable family over an alphabet of size 5 that additionally satisfies a strong monotonicity condition with respect to a fixed order on the alphabet, proving that monotonicity alone cannot salvage testability in the non-Boolean setting.

The technical contributions include developing a “basic form” normalization for formulas, recursive analysis techniques for tree-structured properties, and sophisticated lower bound constructions using communication complexity arguments. The work has implications for computational complexity theory, suggesting that proving property testing lower bounds could be a viable path for proving lower bounds on the representational complexity of functions via read-once formulas. Overall, the paper provides a nearly complete characterization of the testability of read-once formula satisfaction, delineating the precise conditions—Boolean domain, bounded arity, monotonicity—that enable efficient testing and the point where these conditions break down.