Scribd
Upload
60 views

Hwsoln 08

This document contains solutions to 4 problems about decidability and Turing machines. Problem 1 shows that the problem of determining if a DFA and regular expression are equivalent is decidable. Problem 2 shows that the problem of determining if a context-free grammar generates the empty string is decidable. Problem 3 shows that the problem of determining if a regular expression generates a string containing "111" as a substring is decidable. Problem 4 shows that the emptiness problem for Turing machines is co-Turing-recognizable.

Uploaded by

tghate91
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
60 views

Hwsoln 08

This document contains solutions to 4 problems about decidability and Turing machines. Problem 1 shows that the problem of determining if a DFA and regular expression are equivalent is decidable. Problem 2 shows that the problem of determining if a context-free grammar generates the empty string is decidable. Problem 3 shows that the problem of determining if a regular expression generates a string containing "111" as a substring is decidable. Problem 4 shows that the emptiness problem for Turing machines is co-Turing-recognizable.

Uploaded by

tghate91
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 3

CS 341: Foundations of Computer Science II

Prof. Marvin Nakayama

Homework 8 Solutions

1. Consider the decision problem of testing whether a DFA and a regular expression are
equivalent. Express this problem as a language and show that it is decidable.

Answer: Define the language as

C = {hM, Ri | M is a DFA and R is a regular expression with L(M ) = L(R) }.

Recall that the proof of Theorem 4.5 defines a Turing machine F that decides the
language EQDFA = { hA, Bi | A and B are DFAs and L(A) = L(B) }. Then the
following Turing machine T decides C:

T = “On input hM, Ri, where M is a DFA and R is a regular expression:


1. Convert R into a DFA DR using the algorithm in the
proof of Kleene’s Theorem.
2. Run TM decider F from Theorem 4.5 on input hM, DR i.
3. If F accepts, accept. If F rejects, reject.”

2. Consider the decision problem of testing whether a CFG generates the empty string.
Express this problem as a language and show that it is decidable.

Answer: The language of the decision problem is

AεCFG = { hGi | G is a CFG that generates ε }.

If a CFG G = (V, Σ, R, S) includes the rule S → ε, then clearly G can generate


ε. But G could still generate ε even if it doesn’t include the rule S → ε. For
example, if G has rules S → XY , X → aY | ε, and Y → baX | ε, then the derivation
S ⇒ XY ⇒ εY ⇒ εε = ε shows that ε ∈ L(G), even though G doesn’t include
the rule S → ε. So it’s not sufficient to simply check if G includes the rule S → ε to
determine if ε ∈ L(G).
But if we have a CFG G′ = (V ′ , Σ, R′ , S ′ ) that is in Chomsky normal form, then G′
generates ε if and only if S ′ → ε is a rule in G′ . Thus, we first convert the CFG G
into an equivalent CFG G′ = (V ′ , Σ, R′ , S ′ ) in Chomsky normal form. If S ′ → ε is
a rule in G′ , then clearly G′ generates ε, so G also generates ε since L(G) = L(G′ ).
Since G′ is in Chomsky normal form, the only possible ε-rule in G′ is S ′ → ε, so the
only way we can have ε ∈ L(G′ ) is if G′ includes the rule S ′ → ε in R. Hence, if

1
G′ does not include the rule S ′ → ε, then ε 6∈ L(G′ ). Thus, a Turing machine that
decides AεCFG is as follows:
M = “On input hGi, where G is a CFG:
1. Convert G into an equivalent CFG G′ = (V ′ , Σ, R′ , S ′ )
in Chomsky normal form.
2. If G′ includes the rule S ′ → ε, accept. Otherwise, reject.”
3. Let Σ = {0, 1}, and consider the decision problem of testing whether a regular
expression with alphabet Σ generates at least one string w that has 111 as a substring.
Express this problem as a language and show that it is decidable.
Answer: The language of the decision problem is
A = { hRi | R is a regular expression describing a language over Σ containing
at least one string w that has 111 as a substring
(i.e., w = x111y for some x and y) }.
Define the language C = { w ∈ Σ∗ | w has 111 as a substring }. Note that C is a
regular language with regular expression (0 ∪ 1)∗ 111(0 ∪ 1)∗ and is recognized by
the following DFA DC :

0 0, 1

1
1 1
1 2 3 4
0
0

Now consider any regular expression R with alphabet Σ. If L(R) ∩ C 6= ∅, then R


generates a string having 111 as a substring, so hRi ∈ A. Also, if L(R) ∩ C = ∅,
then R does not generate any string having 111 as a substring, so hRi 6∈ A. By
Kleene’s Theorem, since L(R) is described by regular expression R, L(R) must be a
regular language. Since C and L(R) are regular languages, C ∩ L(R) is regular since
the class of regular languages is closed under intersection, as was shown in an earlier
homework. Thus, C ∩L(R) has some DFA DC∩L(R) . Theorem 4.4 shows that EDFA =
{ hBi | B is a DFA with L(B) = ∅ } is decidable, so there is a Turing machine H that
decides EDFA . We apply TM H to hDC∩L(R) i to determine if C ∩ L(R) = ∅. Putting
this all together gives us the following Turing machine T to decide A:
T = “On input hRi, where R is a regular expression:
1. Convert R into a DFA DR using the algorithm in the
proof of Kleene’s Theorem.
2. Construct a DFA DC∩L(R) for language C ∩ L(R)
from the DFAs DC and DR .
3. Run TM H that decides EDFA on input hDC∩L(R) i.
4. If H accepts, reject. If H rejects, accept.”

2
4. Consider the emptiness problem for Turing machines:

ETM = { hM i | M is a Turing machine with L(M ) = ∅ }.

Show that ETM is co-Turing-recognizable. (A language L is co-Turing-recognizable if


its complement L is Turing-recognizable.) Note that the complement of ETM is

ETM = { hM i | M is a Turing machine with L(M ) 6= ∅ }.

(Actually, ETM also contains all hM i such that hM i is not a valid Turing-machine
encoding, but we will ignore this technicality.)

Answer: We need to show there is a Turing machine that recognizes ETM , the com-
plement of ETM . Let s1 , s2 , s3 , . . . be a list of all strings in Σ∗ . For a given Turing
machine M , we want to determine if any of the strings s1 , s2 , s3 , . . . is accepted by
M . If M accepts at least one string si , then L(M ) 6= ∅, so hM i ∈ ETM ; if M
accepts none of the strings, then L(M ) = ∅, so hM i 6∈ ETM . However, we cannot
just run M sequentially on the strings s1 , s2 , s3 , . . .. For example, suppose M accepts
s2 but loops on s1 . Since M accepts s2 , we have that hM i ∈ ETM . But if we run M
sequentially on s1 , s2 , s3 , . . ., we never get past the first string. The following Turing
machine avoids this problem and recognizes ETM :

R = “On input hM i, where M is a Turing machine:


1. Repeat the following for i = 1, 2, 3, . . ..
2. Run M for i steps on each input s1 , s2 , . . . , si .
3. If any computation accepts, accept.

You might also like