Authors: Richard Beigel and and Bin Fu
Abstract:
Wilson's (JCSS '85) model of oracle gates provides a framework for
considering reductions whose strength is intermediate between
truth-table and Turing. Improving on a stream of results by Beigel,
Reingold, Spielman, Fortnow, and Ogihara, we prove that PL and PP are
closed under NC1 reductions. This answers an open problem
of Ogihara (FOCS '96). More generally, we show that
NCk+1PP = ACkPP
and
NCk+1PL = ACkPL
for all k ≥ 0. On the other hand, we construct an oracle A
such that NCkPPA ≠
NCk+1PPA for all integers
Slightly weaker than NC1 reductions are Boolean formula reductions. We ask whether PL and PP are closed under Boolean formula reductions. This is a nontrivial question despite NC1 = BF, because that equality is easily seen not to relativize. We prove that PPPlog²n/loglogn-T ⊆ BFPP ⊆ PrTIME(nO(logn)). Because PPPlog²n/loglogn-T is not contained in PP relative to an oracle, we think it is unlikely that PP is closed under Boolean formula reductions. We also show that PL is unlikely to be closed under BF reductions.