Weizmann Logo
Electronic Colloquium on Computational Complexity

Under the auspices of the Computational Complexity Foundation (CCF)

Login | Register | Classic Style



TR18-118 | 20th June 2018 23:05

Hardness of improper one-sided learning of conjunctions for all uniformly falsifiable CSPs



We consider several closely related variants of PAC-learning in which false-positive and false-negative errors are treated differently. In these models we seek to guarantee a given, low rate of false-positive errors and as few false-negative errors as possible given that we meet the false-positive constraint. Bshouty and Burroughs first observed that learning conjunctions in such models would enable PAC-learning of DNF in the usual distribution-free model; in turn, results of Daniely and Shalev-Shwartz establish that learning of DNF would imply algorithms for refuting random k-SAT using far fewer constraints than believed possible. Such algorithms would violate a slight strengthening of Feige's R3SAT assumption, and would violate the RCSP hypothesis of Barak et al. We show here that actually, an algorithm for learning conjunctions in this model would have much more far-reaching consequences: it gives refutation algorithms for all predicates that are falsified by one of the uniform constant strings. To our knowledge, this is the first hardness result of improper learning for such a large class of natural average-case problems with natural distributions.

ISSN 1433-8092 | Imprint