PH P BPP Basic Definitions Language L belongs to class PP if - - PowerPoint PPT Presentation

ph
SMART_READER_LITE
LIVE PREVIEW

PH P BPP Basic Definitions Language L belongs to class PP if - - PowerPoint PPT Presentation

Toda Theorem Part1 PH P BPP Basic Definitions Language L belongs to class PP if exists polynomial NTM such that = > { ( ) 1 } 0 . 5 x L P M x Language L belongs to class BPP if exists


slide-1
SLIDE 1

Toda Theorem Part1

BPP PH ⊆

P ⊕

slide-2
SLIDE 2

Basic Definitions

  • Language L belongs to class PP if exists

polynomial NTM such that

5 . } 1 ) ( { > = ⇔ ∈ x M P L x

  • Language L belongs to class BPP if

exists probabilistic polynomial algorithm such that x from L is accepted with probability more than 0.75 and x not form L is declined with the same probability.

slide-3
SLIDE 3
  • Language L belongs to class

P ⊕ (parity P) if exists polynomial-time non-deterministic Turing machine, such x lies in L is equivalent to that fact, that number of accepting computations is odd(or, exists polynomially checked relation R such that the number of y(R(x, y)= 1 ) is odd.

Class

P ⊕

  • It can be quickly checked that

is complete for

SAT ⊕ P ⊕

slide-4
SLIDE 4

Polynomial hierarchy

Ρ Σ +

= Σ = Σ

i

NP P P P

i 1

P PH

i i

Σ =

U

  • Polynomial Hierarchy
slide-5
SLIDE 5

General Idea of proof

P

BPP NP

P

RP NP

A P A

BPP NP

P

BPP PH

N i ∈ ∀

P i

BPP P

⊆ Σ

  • because because of

Valiant-Vazirani Lemma

  • because Valiant-Vazirani Lemma can be

relativised

  • We will prove, that by the

mathematical induction. It will prove, that

  • Base i =1 is obvious
slide-6
SLIDE 6

Idea of Proof

  • Lemma 2:
  • Lemma 3:
  • Lemma 4:

So, considering this Lemmas being truth we’ll receive

A P A BPP

BPP P

⊆ ⊕

P P

P

⊕ ⊆ ⊕

A A BPP

BPP BPP ⊆

P P BPP P P BPP P BPP P P BPP P k

BPP BPP BPP BPP NP NP P

k

⊕ ⊕ ⊕ ⊕ ⊕ ⊕ ⊕ Σ +

⊆ ⊆ ⊆ ⊆ ⊆ = Σ

1

and the first part of Toda’s Theorem will be proved.

slide-7
SLIDE 7

Lemma 4

  • In definition of BPP we can use, instead of (1-3/4),

where is a polynomially function of input length

  • M uses L as oracle, M is wrong with pr=
  • We can consider, that the lengths of all branches of NTM L

are equal and the number of its accesses to L is equal to l(n) in all branches

  • We can replace every access to oracle L by the branch of

machine N(from the definition of BPP) with the probability

  • f mistake
  • The number of branches, which status will change is equal

to that can be made less than 0.25

A

BPP L ∈

i

p −

2

i

p

) (

2

n e −

) (

2

n e −

) ( ) ( ) (

) 2 1 ( 1 2

n l n i n e − −

− − +

) (

2

n i −

slide-8
SLIDE 8

Lemma 2

  • so exists NTM with oracle accepting

every for odd number of y. R is corresponding relationship

  • So
  • And we need:

A BPP

P L ⊕ ∈

A A

BPP B ∈

n

x } 1 , { ∈

) (

} 1 , {

n

y

γ

A A BPP A BPP A B

BPP BPP P P R ∈ ∈ ∈ ∈

} } 2 ) 2 1 ( )} ( ) , , ( : { :# { |# {

) ( ) (

  • dd

is L z y x z y x L

n n A ς π −

− ≥ Π ∈ = } 2 * 75 . } )} ( ) , , ( : { :# { |# {

) (n A

  • dd

is L z y x y z x L

ς

≥ Π ∈ =

slide-9
SLIDE 9

Lemma 2- end

  • We can consider a table, (for fixed x, WLOG x is in

L), in (y, z) we’ll put result of П for given (y, z).The number of rows with more than

  • nes is odd and the number of columns, which have

1 in intersection with this columns is more than

  • In other rows is not many 1,so the number of ‘good’

rows is for

) ( ) ( ) ( ) (

2 * 2 * 2 2

n n n n γ ς π ς −

} 2 ) 2 1 (

) ( ) ( n n ς π −

) ( ) ( ) ( ) ( ) ( ) ( ) ( ) (

2 * 75 . 2 * 2 * 2 2 * 2 * 2 2

n n n n n n n n ς γ ς π γ ς π ς

≥ − −

− −

3 ) ( ) ( + ≥ n n γ π

slide-10
SLIDE 10

Lemma 5

P co P ⊕ − = ⊕