In computational complexity theory, SP
2 is a complexity class, intermediate between the first and second levels of the polynomial hierarchy. A language L is in if there exists a polynomial-time predicate P such that
- If , then there exists a y such that for all z, ,
- If , then there exists a z such that for all y, ,
where size of y and z must be polynomial of x.
Relationship to other complexity classes
It is immediate from the definition that SP
2 is closed under unions, intersections, and complements. Comparing the definition with that of and , it also follows immediately that SP
2 is contained in . This inclusion can in fact be strengthened to ZPPNP.
Every language in NP also belongs to SP
2. For by definition, a language L is in NP, if and only if there exists a polynomial-time verifier V(x,y), such that for every x in L there exists y for which V answers true, and such that for every x not in L, V always answers false. But such a verifier can easily be transformed into an SP
2 predicate P(x,y,z) for the same language that ignores z and otherwise behaves the same as V. By the same token, co-NP belongs to SP
2. These straightforward inclusions can be strengthened to show that the class SP
2 contains MA (by a generalization of the Sipser–Lautemann theorem) and (more generally, ).
A version of Karp–Lipton theorem states that if every language in NP has polynomial size circuits then the polynomial time hierarchy collapses to SP
2. This result yields a strengthening of Kannan's theorem: it is known that SP
2 is not contained in Template:San-serif(nk) for any fixed k.
As an extension, it is possible to define as an operator on complexity classes; then . Iteration of operator yields a "symmetric hierarchy"; the union of the classes produced in this way is equal to the Polynomial hierarchy.
- Canetti, Ran (1996). "More on BPP and the polynomial-time hierarchy". Information Processing Letters. 57 (5). Elsevier. pp. 237–241. doi:10.1016/0020-0190(96)00016-6.
- Russell, Alexander; Sundaram, Ravi (1998). "Symmetric alternation captures BPP". computational complexity. 7 (2). Birkhäuser Verlag. pp. 152–162. doi:10.1007/s000370050007. ISSN 1016-3328.