Additive-error PAC-verification for AC^0[2]
Develop PAC-verification protocols for agnostic learning of AC^0[2] under the uniform distribution that achieve additive error guarantees of the form opt(f, AC^0[2]) + ε, potentially by exploiting the algebraic structure arising from Razborov–Smolensky lower bounds.
References
Another question is whether we can use the algebraic structure of AC0[2] arising from to get better PAC-verification protocols that output a hypothesis with just an additive error, i.e., an error of $opt(f,AC0[2]) + \varepsilon$.
— On the Power of Interactive Proofs for Learning
(2404.08158 - Gur et al., 11 Apr 2024) in Future Directions