Difference between revisions of "Complexity Zoo:O"

From Complexity Zoo
Jump to navigation Jump to search
 
(One intermediate revision by one other user not shown)
Line 72: Line 72:
 
<li>If the answer is 'no,' then for all y, P(x,y,z*) is false.</li>
 
<li>If the answer is 'no,' then for all y, P(x,y,z*) is false.</li>
 
</ol>
 
</ol>
Note that this differs from [[#s2p|S<sub>2</sub>P]] in that the witnesses in each case must only depend on the length of the input, and not on the input itself. In particular, since the conditions imply that the answer is 'yes' iff P(x,y*,z*), the class O<sub>2</sub>P is included in [[Complexity Zoo:P#ppoly|P/poly]].
+
Note that this differs from [[Complexity Zoo:S#s2p|S<sub>2</sub>P]] in that the witnesses in each case must only depend on the length of the input, and not on the input itself. In particular, since the conditions imply that the answer is 'yes' iff P(x,y*,z*), the class O<sub>2</sub>P is included in [[Complexity Zoo:P#ppoly|P/poly]].
  
 
Less formally, O<sub>2</sub>P is the class of one-round games in which a prover and a disprover submit simultaneous moves to a deterministic, polynomial-time referee, and furthermore, there is a single winning move the prover can make that works for all x of length n that are yes-instances, and there is a single winning move the disprover can make that works for all x of length n that are no-instances.
 
Less formally, O<sub>2</sub>P is the class of one-round games in which a prover and a disprover submit simultaneous moves to a deterministic, polynomial-time referee, and furthermore, there is a single winning move the prover can make that works for all x of length n that are yes-instances, and there is a single winning move the disprover can make that works for all x of length n that are no-instances.
Line 80: Line 80:
 
Contains [[Complexity Zoo:O#onp|ONP]] and coONP [[zooref#gm15|[GM15]]].
 
Contains [[Complexity Zoo:O#onp|ONP]] and coONP [[zooref#gm15|[GM15]]].
  
Contrast with [[Complexity Zoo:S#s2p|S<sub>2</sub>P]].
+
It follows [[zooref#glv24|[GLV24]]] from results of [[zooref#li23|[Li23]]] that for each ''k'', O<sub>2</sub>P contains a language that has circuit complexity at least ''n<sup>k</sup>''. (Previously, this was known for NP ∪ O<sub>2</sub>P based on [[zooref#cr06|[CR06]]].)

Latest revision as of 20:06, 9 May 2024

Back to the Main Zoo - Complexity Garden - Zoo Glossary - Zoo References


Complexity classes by letter: Symbols - A - B - C - D - E - F - G - H - I - J - K - L - M - N - O - P - Q - R - S - T - U - V - W - X - Y - Z

Lists of related classes: Communication Complexity - Hierarchies - Nonuniform


OIP - OMA - ONP - OptP - O2P


OIP: Oblivious IP

IP where only the input size is known during the interaction with the prover, and after that interaction, the verifier gets the specific input.

L is in OIP if there exists a randomized, polynomial time interrogator I which takes an input size and interacts with a prover to produce a witness, and polynomial time verifier V that takes an input and a witness so that

  1. Given an input size n, there is prover so that I given n and interacting with the prover produces a witness w so that, for any input x of length n where x is in L, V accepts on x and w.
  2. For any x not in L of length n, for any prover interacting with I on input n gives a witness w' which V will reject with x and w' with probability at least 2/3.

OIP = IPP/poly [GM15].


OMA: Oblivious MA

The class of functions computable in randomized polynomial time with a shared, untrusted witness for each input size. The input-oblivious version of MA.

L is in OMA if there exists a randomized, polynomial time verifier V taking an input and a witness, so that:

  1. There is a witness for each n of polynomial size, so that for any input of size n, if the input is is in L, then the verifier accepts on that input and the witness.
  2. If the input is not in L, then for any witness, the verifier rejects on that input with probability at least 1/2.

NP is contained in OMA iff NP is in P/poly [FSW09].

EXP is contained in P/poly iff EXP = OMA [FSW09].

BPP is contained in OMA [GM15].

Implicit in [San07] that OMA/1 (that is, OMA with 1 bit of trusted advice) does not have circuits of size for any .


ONP: Oblivious NP

The class of functions computable in polynomial time with a shared, untrusted witness for each input size. The input-oblivious version of NP.

L is in ONP if there exists a polynomial time verifier V taking an input and a witness, so that:

  1. There is a witness for each n of polynomial size, so that for any input of size n, if the input is in L, then the verifier accepts on that input and the witness.
  2. If the input is not in L, then for any witness, the verifier rejects on that input.

Defined in [FSW09], where it was shown NP has size nk circuits for some constant k if and only if ONP/1 has size nj circuits for some constant j.

ONP is contained in P/poly and NP [FSW09].

ONP = NP iff NP is in P/poly [FSW09].

If NE is not E then ONP is not P [GM15].

See also YP for an input oblivious analogue of NP ∩ coNP.


OptP: Optimum Polynomial-Time

The class of functions computable by taking the maximum (or minimum) of the output values over all accepting paths of an NP machine. OptP[f(n)] restricts the size of the output to f(n) bits.

Defined in [Kre88].

Contrast with FNP.

MaxSat that returns the number of satisfied clauses, Clicque that returns the size of the clique, Chromatic Number are OptP[log n]-complete.


O2P: Second Level of the Oblivious Symmetric Hierarchy

The class of decision problems for which there is a polynomial-time predicate P such that, for each length n, there exists y* and z* of length poly(n) such that for all x of length n

  1. If the answer is 'yes,' for all z, P(x,y*,z) is true.
  2. If the answer is 'no,' then for all y, P(x,y,z*) is false.

Note that this differs from S2P in that the witnesses in each case must only depend on the length of the input, and not on the input itself. In particular, since the conditions imply that the answer is 'yes' iff P(x,y*,z*), the class O2P is included in P/poly.

Less formally, O2P is the class of one-round games in which a prover and a disprover submit simultaneous moves to a deterministic, polynomial-time referee, and furthermore, there is a single winning move the prover can make that works for all x of length n that are yes-instances, and there is a single winning move the disprover can make that works for all x of length n that are no-instances.

Defined in [CR06], where it was shown (among other properties) that O2P is self-low, and that the Karp–Lipton collapse goes all the way down to O2P: if NP is contained in P/poly then PH = O2P.

Contains ONP and coONP [GM15].

It follows [GLV24] from results of [Li23] that for each k, O2P contains a language that has circuit complexity at least nk. (Previously, this was known for NP ∪ O2P based on [CR06].)