
Nonrelativizing Separations Harry Buhrman* Lance Fortnow t Thomas Thierauf+ CWI Dept. of Computer Science Abt. Theoretische Informatik University of Chicago Universitat Ulm PO Box 94079 1100 E. 58th St. Oberer Eselsberg 1090 GB Amsterdam Chicago, IL 60637 89069 Ulm The Netherlands USA Germany Abstract help us decide where and how to put our efforts into solving We show that MAExP. the exponential time version of problems in complexity theory. It is still true that virtually the Merlin-Arthur class, does not have polynomial size cir­ all of the theorems in computational complexity theory that cuits. This significantly improves the previous known result have reasonable relativizations do relativize (see [For94]). due to Kannan since we furthemzore show that our result But we do have a small number of exceptions that arise does not relativize. This is the first separation result in com­ from the area of interactive proofs. These results have pre­ plexity theory that does not relativize. As a corollary to our viously always taken the form of collapses such as IP = separation result we also obtain that PEXP, the exponen­ PSPACE [LFKN92, Sha92], MIP = NEXP [BFL91] and tial time version of PP is not in P /poly. PCP(O(l), O(logn)) =NP [ALM+92]. In this paper we give the first reasonable nonrel­ 1 Introduction ativizing separation results. Namely, we show that Baker, Gill and Solovay [BGS75] noticed that the re­ there exist languages in MAExP, a one-round Merlin­ sults proven in computational complexity theory relativize, Arthur game [BM88] with an exponential-time verifier, i.e. the proofs go through virtually unchanged if all of the that cannot have polynomial-size circuits, in other words, machines involved have access to the same information via MAExP ~ P/poly. On the other hand, we then cre­ an oracle. They then developed relativized worlds where ate a relativized world where every MAExP language has P = NP and P =f. NP and thus argued that the current tech­ polynomial-size circuits, i.e., MAExP ~ P/poly. niques in complexity theory could not settle this question. Paul, Pippenger, Szemeredi and Trotter [PPST83] show More than two decades later, relativization still plays a separation of nondeterministic linear time from determin­ in important role in complexity theory. Though no longer istic linear time where one can create a relativized world believed to be any fonn of an independence result, they do where these two classes coincide. However, both the sepa­ ration and the oracle heavily depend on the machine model *Email: [email protected]. URL: http://www.cwi.nl!buhrrnan. Par­ where our results are model independent. tially supported by the Dutch foundation for scientific research (NWO) by SION project 612-34-002, and by the European Union through Neu­ We can strengthen our oracle to work for MIPExP, the roCOLT ESPRIT Working Group Nr. 8556, and HC&M grant nr. languages accepted by multiple prover interactive proof ERB4050PL93-05 I 6. systems [BGKW88] with an exponential-time verifier. In t Email: [email protected]. URL: http://www.cs.uchicago.edurfortnow. Work done while on leave at fact, besides having small circuits, relative to our oracle CWI. Email: [email protected]. Supported in part by NSF grant MIPExP languages can be accepted in pNP and in EBP. CCR 92-53582, the Dutch Foundation for Scientific Research (NWO) and Since these classes are contained in MIPExp, we get a rel- a Fulbright Scholar award. i Email: [email protected]. URL: http://hermes.infonnatik.uni-ulm.de/ti/Personen/tt.html. 8 1093-0159/98 $10.00 © 1998 IEEE ativized world where Let # M ( x) represent the number of accepting compu­ tations of a nondeterministic Turing machine M ( x ), and P/poly n pNP n EBP. #MR( x) the number of rejecting computations of M ( x ). This contrasts greatly with the situation in the unrelativized A language L is in EBP if there exists a polynomial-time world where we have nondeterministic Turing machine M such that for all x, pNP U 4P C PSPACE x EL {:=::} #M(x) is odd. C EXPSPACE A language L is in PP if there exists a polynomial-time C NEEXP nondetenninistic Turing machine M such that for all x, MIPExP, x EL {:=::} #M(x) > #MR(x). where NEEXP is nondetenninistic double exponential PEXP is the exponential-time version of PP, i.e., the time. polynomial-time nondeterministic Turing machine in the Our proof that MAExP does not lie in P /poly uses the definition of PP is nondeterministic exponential-time. result of Babai, Fortnow, Nisan and Wigderson [BFNW93] A language L is in MA if there exists a probabilistic that if EXP has polynomial-size circuits then EXP= MA. polynomial time Turing machine M and a polynomial q( n) Since this is the only part of the proof that does not rel­ such that ativize, our paper gives the first oracle where this Babai­ Fortnow-Nisan-Wigderson result does not relativize. • x EL=> 3y E ~q(lxlJ Pr(M(x, y) accepts) 2:: 2/3, As a corollary to our separation result we also obtain the and separation between PEXP, the exponential time version of I:;q(lxll Pr(M(x, y) accepts):::; 1/3. PP, and P/poly. Finding nonrelativizing results like these • x ~ L => \fy E allows us to better understand the importance and limita­ This corresponds to the Merlin-Arthur games due to Babai tions of relativization and gives us new ideas on how to and Moran [BM88]. prove other nonrelativizing results. We also consider multiple-prover interactive proof sys­ 2 Preliminaries tems as developed by Ben-Or, Goldwasser, Kilian and the verifier can ask We use (J:1, ... , .tk) to be any polynomial-time com­ Wigderson [BGKW88). In this model, are unable to communi­ putable and invertible tupling function. For a language A, questions of several provers that cate with each other. Babai, Fortnow and Lund [BFL9 l] we denote the characteristic function of A as A(.). that the class, MIP, of languages provable by such We assume the reader familiar with basic notations in show is equal to NEXP. complexity theory and classes such as P and NP. We let systems 0111 0111 We use MAExP and MIPExP to represent the Merlin­ EXP = DTIME[~" ] and NEXP = NTIME[2" ]. Arthur games and multiple-prover interactive proof sys­ The class P/poly consists of languages accepted by tems where we allow the verifier to use time 2n • for some a family of polynomial-size circuits or equivalently a fixed k. In particular, the provers can send messages up to polynomial-time Turing machine given a polynomially­ this length to the verifier. length advice that depends only on the length of the input. More fonnally, L E P/poly if there exist A E Panda 3 A Nonrelativizable Separation polynomially length bounded function h : N r-. E* such The computational power of polynomial-size circuits, that for al I i: P /poly, is an interesting issue. In particular, whether one can solve all sets in NP within P /poly is still an open ques­ x EL -<::::==> (x,h(jxj)) EA. tion, although there are strong indications that this is not The value h(IJ:I) is called the advice for strings of length possible: otherwise the polynomial hierarchy collapses to 1x1. I:~ [KL80]. 9 With respect to absolute separations, Kannan [Kan82] But this contradicts Theorem 3.1. 0 that there are sets in NEXPNP that cannot be com­ showed The same proof works for MAExP n coMAExP instead by polynomial-size circuits. puted of just MAEXP· Then we need the form of Kannan's result Theorem 3.1 (Kannan) as stated in Theorem 3.1. NEX~P n coNEXPNP </;. P /po! y. Corollary 3.5 MAExP n coMAEXr i P /poly. We improve Kannan 's result and show that there are sets Vereshchagin [Ver92] shows that MA is in PP. By in MAExP that cannot be computed by polynomial-size cir­ padding analogously to the proof of Lemma 3.3 this im­ cuits. We use the following result ofBabai, Fortnow, Nisan plies that MAExP ~ PEXP. and Wigderson [BFNW93]. Corollary 3.6 PEXP i P/poly. Theorem 3.2 (BFNW) EXP~ P/poly => EXP= MA. 4 Relativized Collapse In order to prove the separation result we will also need In this section we show that our separation result from the following lemma. the previous section does not relativize. This is the first known example of a non-relativizing separation result. Lemma3.3 N~P = MA => NEX~P = MAEXP· Theorem 4.1 There exists an oracle A such that Proof: Let A be a set in NEX~P, and let this be MA~xp ~ pA /poly. witnessed by an alternating Turing machine that runs in Proof: Let /I.Ji be an enumeration of potential verifiers, time 2P(n) for some polynomial p. Consider the following i.e., probabilistic Turing machines, that run in time 2n. It is padded version of A: sufficient to encode just these machines: for verifiers that A'= { (x, 02 v(lxl)) IX EA}. use time 2n k we can use padding to reduce the language to a verifier that uses 2" time. It follows that A' is in N~P and by assumption is in MA. We first describe how to encode inputs of a single This however implies that removing the padding yields that length. Later we will show how to combine lengths. For 0 A E MAEXP· inputs of length n, we will encode languages proven to ver­ We are now ready to prove our separation result. ifiers M1, ..
Details
-
File Typepdf
-
Upload Time-
-
Content LanguagesEnglish
-
Upload UserAnonymous/Not logged-in
-
File Pages5 Page
-
File Size-