BPP Contained in PH
By Michael Sipser;
Clemens Lautemann
Presenter: Jie Meng
Sipser-Lautemann Theorem
• M. Sipser. A complexity theoretic
approach to randomness, In
Proceedings of the 15th ACM STOC, 1983
• C. Lautemann, BPP and the polynomial
hierarchy, Information Process Letter 14
215-217, 1983
Sipser-Lautemann Theorem
Outline
•
•
•
•
•
Definition and Background
Techniques
Proof
Questions
Homework
Sipser-Lautemann Theorem
BPP: A language L is in BPP if and only if
there exists a randomized Turing Machine M,
s.t.
2
x L Prr [ M ( x, r ) 1]
3
1
x L Prr [ M ( x,r) 1]
3
Sipser-Lautemann Theorem
Trivially
P RP BPP PP PSPACE
P RP NP PP PSPACE
BPP BPP
But, BPP ? NP
Sipser-Lautemann Theorem
Main Theorem:
BPP PH
Actually BPP 2p 2p
Sipser-Lautemann Theorem
Let C be a language, NPC be the class
that L is in NPC if there is a nondeterministic Turing Machine M, which
can accept L, with the power that M can
query an oracle such questions like “if y
is in C” and get the correct answer in one
step.
This can be generalized to NPA, A is a
language class.
Sipser-Lautemann Theorem
NP: L is in NP if there exists a
deterministic polynomial Turing Machine
M, s.t. xL y M(x,y)=1
Sipser-Lautemann Theorem
L in NP, for any x in L
Rej Rej Rej Rej Acc Rej Rej Rej Acc Rej Rej Acc Rej Rej Rej Rej
Sipser-Lautemann Theorem
L in NP, for any x not in L
Rej Rej Rej Rej Rej Rej Rej Rej Rej Rej Rej Rej Rej Rej Rej Rej
Sipser-Lautemann Theorem
def
def
1 NP
def
i 1 NP
def
i co i
i
PH i
i 1
def
i 1 P
i
Sipser-Lautemann Theorem
L in 2p , x in L, L’ in NP
y in L’ ?
Yes/No
Rej Rej Rej Rej Acc Rej Rej Rej Acc Rej Rej Acc Rej Rej Rej Rej
Sipser-Lautemann Theorem
L in 2p , x in L,
y in L’
Rej Rej Rej Rej Acc Rej Rej Rej Rej Rej Rej Rej Acc Rej Rej Rej
Sipser-Lautemann Theorem
L in 2p , x in L,
y not in L’
Rej Rej Rej Rej Acc Rej Rej Rej Rej Rej Rej Rej Acc Rej Rej Rej
Sipser-Lautemann Theorem
Rej Rej Rej Rej Acc Acc Acc Acc Rej Rej Rej Rej Rej Acc Rej Rej
Sipser-Lautemann Theorem
Equivalent definition
NP: L is in NP, if and only if there exists a
deterministic poly-time TM M, s.t.
xL y M(x,y)=1
2p: L is in 2p , if and only if there exists a
deterministic poly-time TM M’, s.t.
x L y z M’(x,y,z)=1
x L y z M’(x,y,z)=0
Sipser-Lautemann Theorem
Technique
fat
325 lbs
7’ 1’’
VS
Thin
Sipser-Lautemann Theorem
Technique
Sipser-Lautemann Theorem
Technique
Sipser-Lautemann Theorem
PROOF
L in BPP:
2
x L Prr [ M ( x, r ) 1]
3
1
x L Prr [ M ( x,r) 1]
3
By amplifying method and Chernoff Bound
1
x L Prr{0,1}m [ M ( x , r ) 1] 1
m
1
x L Prr{0,1}m [ M ( x , r ) 1]
m
Sipser-Lautemann Theorem
PROOF
Wx={ y | M(x, y)=1}
x in L, |Wx|>2m (1-1/m),
Wx is very fat;
x not in L, |Wx|<2m 1/m
Wx is very thin;
{0,1}m is the whole space;
Sipser-Lautemann Theorem
PROOF
Shifting:
y {0,1}m ,
Wx y {z | t Wx, t y z}
If Wx is fat, |Wx|>2m (1-1/m),
There exists a set of strings y1, y2, … yr, r=m/2
s.t.
yi Wx yi {0, 1}m
If Wx is thin, |Wx|<2m 1/m
There is no such set of strings
2m
1
r 2m
m
Sipser-Lautemann Theorem
PROOF
X in L, Wx is fat,
there exists a set of string y1, y2, … yr
yi Wx yi {0, 1}
m
Then for all z in {0, 1}m ,
z yi Wx y i
That is, there exists i, s.t.
z yi Wx
Sipser-Lautemann Theorem
PROOF
x in L, Wx is fat,
There exists y1, y2, … yr,
For all z in {0,1}m, M(x, z yi)=1, for some i;
x in L,
y1 y 2 ... y r z
M(x, y1 z) 1 or M(x, y 2 z) 1 or ... or M(x, y r z) 1
Sipser-Lautemann Theorem
Question?
Sipser-Lautemann Theorem
HOMEWORK
Finish
the proof in case x is not in L, which
is to say, fill out the blank in the following
statement:
L in BPP, x L […] […]
M’(x,y,z)=[.]
Give all necessary explanations about
your statement.
© Copyright 2026 Paperzz