MÔ HÌNH MÁY TURING

(Turing Machines Model) Definition:

2
A language is recursively enumerable
Recursively Enumerable
if some Turing machine accepts it
and
Recursive Languages
« If the human brain was so simple
that we could understand it,
then we would be so simple that we could not »
(L. Watson)
2/96

Let L be a recursively enumerable language

Definition:
and M the Turing Machine that accepts it
A language is recursive
For string w: if some Turing machine accepts it
if w∈ L then M halts in a final state and halts on any input string

In other words:
if w∉ L then M halts in a non-final state
A language is recursive if there is
or loops forever a membership algorithm for it
3/96 4/96

Let L be a recursive language We will prove:

and M the Turing Machine that accepts it 1. There is a specific language

which is not recursively enumerable
For string w: (not accepted by any Turing Machine)

if w∈ L then M halts in a final state 2. There is a specific language

which is recursively enumerable
if w ∉ L then M halts in a non-final state but not recursive
5/96 6/96

1
 Non Recursively Enumerable

A Language which
Recursively Enumerable
is not
Recursively Enumerable
Recursive

7/96

We want to find a language that Consider alphabet {a}

is not Recursively Enumerable
Strings: a, aa, aaa, aaaa, K
This language is not accepted by any
Turing Machine a1 a 2 a3 a4 L

9/96 10/96

Example language accepted by M i

Consider Turing Machines

L( M i ) = {aa, aaaa, aaaaaa}
that accept languages over alphabet {a}
L( M i ) = {a 2 , a 4 , a 6 }

Alternative representation
They are countable:

M1, M 2 , M 3 , M 4 , K a1 a2 a3 a4 a5 a6 a7 L

L( M i ) 0 1 0 1 0 1 0 L
11/96 12/96

2
 Consider the language
a1 a2 a3 a4 L

L ( M1 ) 0 1 0 1 L L = {a i : a i ∈ L( M i )}
L( M 2 ) 1 0 0 1 L

L( M 3 ) 0 1 1 1 L
L consists from the 1’s in the diagonal
L( M 4 ) 0 0 0 1 L
13/96 14/96

a1 a2 a3 a4 L Consider the language L

L(M 1) 0 1 0 1 L
L = {a i : a i ∈ L( M i )}
L(M 2 ) 1 0 0 1 L

L
L(M 3 ) 0 1 1 1
L = {a i : a i ∉ L( M i )}
L(M 4 ) 0 0 0 1 L
L consists of the 0’s in the diagonal
L = {a 3 , a 4 ,K} 15/96 16/96

a1 a2 a3 a4 L
Theorem:
L(M 1) 0 1 0 1 L
Language Lis not recursively enumerable
L(M 2 ) 1 0 0 1 L

L(M 3 ) 0 1 1 1 L

L(M 4 ) 0 0 0 1 L

L = {a1, a 2 ,K} 17/96 18/96

3
 Proof: a1 a2 a3 a4 L

L(M 1) L
Assume for contradiction that 0 1 0 1

L is recursively enumerable L(M 2 ) 1 0 0 1 L

L(M 3 ) 0 1 1 1 L
There must exist some machine Mk
L(M 4 ) 0 0 0 1 L
that accepts L

L( M k ) = L 19/96
Question: M k = M1 ? 20/96

a1 a2 a3 a4 L
a1 a2 a3 a4 L
L(M 1) 0 1 0 1 L
L(M 1) 0 1 0 1 L
L(M 2 ) 1 0 0 1 L
L(M 2 ) 1 0 0 1 L
L(M 3 ) 0 1 1 1 L
L(M 3 ) 0 1 1 1 L

L(M 4 ) 0 0 0 1 L
L(M 4 ) 0 0 0 1 L
a1 ∈ L( M k )
Answer: M k ≠ M1 Question: M k = M 2 ?
a1 ∉ L( M1 ) 21/96 22/96

a1 a2 a3 a4 L a1 a2 a3 a4 L
L(M 1) 0 1 0 1 L L(M 1) 0 1 0 1 L
L(M 2 ) 1 0 0 1 L 1 0 0 1 L
L(M 2 )
L(M 3 ) 0 1 1 1 L L(M 3 ) 0 1 1 1 L

L(M 4 ) 0 0 0 1 L L(M 4 ) 0 0 0 1 L
2
a ∈ L( M k )
Answer: Mk ≠ M2 Question: M k = M 3 ?
a 2 ∉ L( M 2 ) 23/96 24/96

4
 a1 a2 a3 a4 L
Similarly: M k ≠ Mi for any i
L(M 1) 0 1 0 1 L

L(M 2 ) 1 0 0 1 L Because either:

L(M 3 ) 0 1 1 1 L
a i ∈ L( M k ) a i ∉ L( M k )
or
L(M 4 ) 0 0 0 1 L
3 a ∉ L( M i )
i
a i ∈ L( M i )
a ∉ L( M k )
Answer: M k ≠ M3
a3 ∈ L( M 3 ) 25/96 26/96

Observation:
Therefore, the machine M k cannot exist
There is no algorithm that describes L
Therefore, the language L
is not recursively enumerable
(otherwise L would be accepted by
some Turing Machine)
End of Proof

27/96 28/96

Non Recursively Enumerable

A Language which is
Recursively Enumerable L Recursively Enumerable
and not Recursive

Recursive

29/96

5
 We want to find a language which

Is recursively But not We will prove that the language

enumerable recursive
L = {a i : a i ∈ L( M i )}

There is a Is recursively enumerable

The machine
Turing Machine but not recursive
doesn’t halt
that accepts on some input
the language
31/96 32/96

a1 a2 a3 a4 L Theorem:
L(M 1) 0 1 0 1 L

1 0 0 1 L
The language L = {a i : a i ∈ L( M i )}
L(M 2 )

L(M 3 ) 0 1 1 1 L is recursively enumerable

L(M 4 ) 0 0 0 1 L

L = {a 3 , a 4 ,K} 33/96 34/96

Turing Machine that accepts L

For any input string w
Proof:
• Compute i , for which w = ai
• Find Turing machine Mi
We will give a Turing Machine that
(using an enumeration procedure
accepts L
for Turing Machines)

• Simulate M i on input a i
• If M i accepts, then accept w
35/96 End of Proof 36/96

6
 Observation:

Recursively enumerable Theorem:

L = {a i : a i ∈ L( M i )}
The language L = {a i : a i ∈ L( M i )}
is not recursive
Not recursively enumerable
L = {a i : a i ∉ L( M i )}
(Thus, also not recursive)
37/96 38/96

Proof: Therefore:
Assume for contradiction that is
L
recursive
L is recursive
Then L is recursive:
Take the Turing Machine M that accepts L But we know:
M halts on any input: L is not recursively enumerable
If M accepts then reject thus, not recursive
If M rejects then accept CONTRADICTION!!!!
39/96 40/96

Non Recursively Enumerable L

Therefore, L is not recursive
Recursively Enumerable

L
Recursive

End of Proof 41/96 42/96

7
 We will prove:
(weak result)
Turing acceptable languages • If a language is recursive then
and
Enumeration Procedures there is an enumeration procedure for it
(strong result)
• A language is recursively enumerable
if and only if
there is an enumeration procedure for it
44/96

Proof:
Theorem: Enumeration Machine

if a language L is recursive then

~ M
there is an enumeration procedure for it M

Enumerates all
Accepts L
45/96 strings of input alphabet 46/96

Enumeration procedure
If the alphabet is {a, b} then
~ Repeat:
M can enumerate strings as follows:
a ~
M generates a string w
b
aa M checks if w∈ L
ab
ba YES: print w to output
bb NO: ignore w
aaa
aab
...... 47/96 End of Proof 48/96

8
 Example: L = {b, ab, bb, aaa,....}
Enumeration
~ L(M ) Theorem:
M Output
a if language L is recursively enumerable then
b b b there is an enumeration procedure for it
aa
ab ab ab
ba
bb bb bb
aaa aaa aaa
aab
...... ...... ...... 49/96 50/96

Proof: If the alphabet is {a, b} then

Enumeration Machine ~
M can enumerate strings as follows:
a
~ M b
M aa
ab
ba
bb
Enumerates all aaa
Accepts L aab
strings of input alphabet 51/96 52/96

NAIVE APPROACH BETTER APPROACH

Enumeration procedure ~
M Generates first string w1
Repeat: M~ generates a string w
M executes first step on w1
M checks if w∈ L
YES: printw to output ~
M Generates second string w2
NO: ignore w
M executes first step on w2
Problem: If w ∉ L second step on w1
machine M may loop forever 53/96 54/96

9
~
M Generates third string w3 w1 w2 w3 w4 L

M executes first step on w3 1 1 1 1

second step on w2 Step
in 2 2 2 2
third step on w1
string
3 3 3 3
And so on............

55/96 L 56/96

Theorem:
If for any string wi
machine M halts in a final state
If for language L
then it prints wi on the output
there is an enumeration procedure
then L is recursively enumerable

End of Proof 57/96 58/96

Proof: Turing machine that accepts L

Input Tape
For input string w
w
Machine that Repeat:
accepts L • Using the enumerator,
generate the next string of L
Enumerator • Compare generated string with w
Compare
for L
If same, accept and exit loop

59/96 End of Proof 60/96

10
 The Chomsky Hierarchy

Non-recursively enumerable
We have proven:
Recursively-enumerable
A language is recursively enumerable Recursive
if and only if
there is an enumeration procedure for it Context-sensitive

Context-free

Regular
61/96 62/96

Consider problems with answer YES or NO

Examples:
Decidability
• Does Machine M have three states ?

• Is string w a binary number?

• Does DFA M accept any input?

64/96

A problem is decidable if some TM

decides (solves) the problem The Turing machine that decides (solves)
Decidable problems: a problem answers YES or NO
for each instance of the problem
• Does Machine M have three states ?

• Is string w a binary number? Input YES

problem Turing Machine
• Does DFA M accept any input?
instance NO
65/96 66/96

11
 Turing Machine that decides a problem

The machine that decides (solves) a problem:

YES states
• If the answer is YES
then halts in a yes state

• If the answer is NO NO states

then halts in a no state

These states may not be final states YES and NO states are halting states
67/96 68/96

Difference between Some problems are undecidable:

Recursive Languages and Decidable problems
which means:
there is no Turing Machine that
solves all instances of the problem
For decidable problems:
A simple undecidable problem:
The YES states may not be final states
The membership problem
69/96 70/96

The Membership Problem Theorem:

The membership problem is undecidable
Input: •Turing Machine M
(there are M and w for which we cannot
•String w decide whether w ∈ L ( M ) )

Question: Does M accept w ?

Proof: Assume for contradiction that
w ∈ L( M ) ? the membership problem is decidable
71/96 72/96

12
 Thus, there exists a Turing Machine H Let L be a recursively enumerable language
that solves the membership problem
Let M be the Turing Machine that accepts L
M YES M accepts w
H
w We will prove that L is also recursive:
NO M rejects w
we will describe a Turing machine that
accepts L and halts on any input
73/96 74/96

Turing Machine that accepts Therefore, L is recursive

L
and halts on any input
Since L is chosen arbitrarily, every
recursively enumerable language is also
H recursive
M YES accept w
M accepts w ? But there are recursively enumerable
w NO reject languages which are not recursive
w

75/96 Contradiction!!!! 76/96

Another famous undecidable problem:

Therefore, the membership problem
is undecidable
The halting problem

END OF PROOF 77/96 78/96

13
 The Halting Problem Theorem:
The halting problem is undecidable
Input: •Turing Machine M
(there are M and w for which we cannot
•String w
decide whether M halts on input w )

Question: Does M halt on input w ?

Proof: Assume for contradiction that
the halting problem is decidable
79/96 80/96

Thus, there exists Turing Machine H Construction of H

that solves the halting problem
H
Input:
M YES M halts on w q y YES
initial tape contents
H
wM w q0
w NO doesn’t
M w qn NO
halt on
Encoding String
of M w
81/96 82/96

Construct machine H′ :
H′
If H returns YES then loop forever
H Loop forever

If H returns NO then halt q y YES qa qb

wM w q0
qn NO

83/96 84/96

14
 Construct machine Ĥ :

Input: wM (machine M ) Ĥ

copy wM wM
If M halts on input wM
wM
wM H′

Then loop forever

Else halt
85/96 86/96

Run machine Ĥ with input itself: Ĥ on input wHˆ :

Input: wHˆ (machine Ĥ ) If Ĥ halts then loops forever

If Ĥ doesn’t halt then it halts

If Ĥ halts on input wHˆ
Then loop forever NONSENSE !!!!!

Else halt
87/96 88/96

Therefore, we have contradiction Another proof of the same theorem:

The halting problem is undecidable

If the halting problem was decidable then
every recursively enumerable language
would be recursive
END OF PROOF

89/96 90/96

15
 Theorem:

The halting problem is undecidable There exists Turing Machine H

that solves the halting problem

Proof:
M YES M halts on w
Assume for contradiction that
the halting problem is decidable H
w NO doesn’t
M w
halt on

91/96 92/96

Turing Machine that accepts L

Let L be a recursively enumerable language
and halts on any input
Let M be the Turing Machine that accepts L M H
NO reject w
We will prove that L is also recursive: M halts on w ?
w YES
we will describe a Turing machine that
accept w
accepts L and halts on any input Halts on final state
Run M
with input w reject w
Halts on non-final
93/96 state 94/96

Therefore L is recursive
Since L is chosen arbitrarily, every
Therefore, the halting problem is undecidable
recursively enumerable language
is also recursive

But there are recursively enumerable

languages which are not recursive

Contradiction!!!!
95/96 END OF PROOF 96/96

16