Scribd
Upload
48 views

CS5371 Theory of Computation: Lecture 12: Computability III (Decidable Languages Relating To DFA, NFA, and CFG)

This document discusses several decidable languages related to DFAs, NFAs, and CFGs. It presents theorems showing that the languages of strings accepted by a DFA (ADFA), NFA (ANFA), regular expression (ARE), and CFG (ACFG) are decidable. It also shows the emptiness and equality problems for DFAs (EDFA, EQDFA) and CFGs (ECFG) are decidable. However, equality of two CFGs (EQCFG) is undecidable. The document concludes by proving any CFL is decidable by a TM simulating the CFG.

Uploaded by

Kamal Walia
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
48 views

CS5371 Theory of Computation: Lecture 12: Computability III (Decidable Languages Relating To DFA, NFA, and CFG)

This document discusses several decidable languages related to DFAs, NFAs, and CFGs. It presents theorems showing that the languages of strings accepted by a DFA (ADFA), NFA (ANFA), regular expression (ARE), and CFG (ACFG) are decidable. It also shows the emptiness and equality problems for DFAs (EDFA, EQDFA) and CFGs (ECFG) are decidable. However, equality of two CFGs (EQCFG) is undecidable. The document concludes by proving any CFL is decidable by a TM simulating the CFG.

Uploaded by

Kamal Walia
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 27

CS5371

Theory of Computation
Lecture 12: Computability III
(Decidable Languages relating
to DFA, NFA, and CFG)
Objectives
•Recall that decidable languages are
languages that can be decided by TM
(that means, the corresponding TM will
accept or reject correctly, never loops)
•In this lecture, we investigate some
decidable languages that are related to
DFA, NFA, and CFG
–Testing Acceptance, Emptiness, or Equality
•Also, we show how TM can simulate CFG
Acceptance by DFA
Let ADFA be the language
{
B, w| B is a DFA that accepts w}
where 
B, wdenotes the encoding of B followed by w

For example, if D is a DFA accepting even length


strings, then,
D, 01,D, 0000are strings in ADFA ,
but D, 1,D, 000are not
Acceptance by DFA (2)
Theorem 1: ADFA is a decidable language

Proof: We construct a TM DADFA that


decides ADFA as follows:
DADFA = “
On input B, w
1. Simulate B on input w
2. If the simulation ends in an accept
state, accept. Else, reject ”
Acceptance by DFA (3)
Q1: How can DADFA perform the above steps??
•It uses 3 tapes; initially, Tape 1 stores the input
B, w, the other two all blanks
•Then, DADFA copies w into Tape 2, and write the
start state of B in Tape 3
•Usage: Tape head of Tape 2 points to next char
in w for B to read, Tape 3 stores current state
•Based on Tapes 2 and 3, DADFA moves back and
forth Tape 1 to know how B performs each
transition, and update the 3 tapes accordingly
Acceptance by DFA (4)
Q2: Why is DADFA a decider for ADFA ?
•For any input 
B, w , it can simulate B so that
each transition in B takes finite number of steps
•To know which state B is at after reading w,
there are only |w| transitions in B
•Thus, it takes finite number of steps to know if
B accepts w or not. So, DADFA can decide (no
infinite loop) whether to accept or reject B, w
Acceptance by NFA
Let ANFA be the language
{
B, w| B is an NFA that accepts w}

Theorem 2: ANFA is a decidable language


Acceptance by NFA (2)

Proof:
[Solution 1] We can use the same idea
when we simulate NTM by TM, so that
we give a TM that decides ANFA.
Precisely, we need to try every possible
branch of computation, but only of
length up to |Q||w| + |Q| (why??)
Acceptance by NFA (3)
[Solution 2 (easier)] We re-use DADFA to
give a TM DANFA that decides ANFA:

DANFA = “
On input B, w
1. Convert B to an equivalent DFA C
2. Run DADFA on C, w
3. If DADFA accepts, accept. Else, reject”
Acceptance by NFA (4)

Q1: How can DANFA perform the above steps??


•It uses 5 tapes; initially, Tape 1 stores the input
B, w
, Tape 2 stores the encoding of DADFA, the
other three all blanks
•Then, DANFA converts B to C and store C in Tape 3
•It then consults DADFA in Tape 2, to know how DADFA
simulates C running on w
•Tapes 4 and 5 can be used to store the current
state of C, and the next char for C to read, so
that DANFA can simulate DADFA to simulate C
Acceptance by NFA (5)

Q2: Why is DANFA a decider for ANFA ?


•For any input 
B, w, it convert B into the
equivalent DFA C in finite number of steps

•Then, it consults DADFA which takes finite


number of steps to know if C accepts w or not.
Thus, DANFA can decide (no infinite loop)
whether to accept or reject  B, w
Acceptance by
Regular Expression (RE)
Let ARE be the language
{
R, w| R is an RE that generates w }

Theorem 3: ARE is a decidable language


Acceptance by RE (2)

Proof: W give a TM DARE that decides ARE:

DARE = “
On input 
R, w
1. Convert R to an equivalent NFA A
2. Run DANFA on 
A, w
3. If DANFA accepts, accept. Else, reject”
Emptiness Test for DFA
Let EDFA be the language
{B| B is a DFA and L(B) = { } }

Theorem 4: EDFA is a decidable language

Observation: DFA accepts no string if and


only if we cannot reach any accept state
from the start state by following
transition arrows
Emptiness Test for DFA (2)
Proof: We use similar idea as we test if a
graph G is connected. Precisely, we give
a TM DEDFA that decides EDFA as follows:
DEDFA = “On input B
1. Mark the start state of B
2. Repeat until no new states are marked
•Mark any state that has a transition
coming into it from a marked state
3. If no accept state of B is marked,
accept. Else, reject”
Equality Test for DFA
Let EQDFA be the language
{A,B| A and B are DFAs and L(A) = L(B) }

Theorem 5: EQDFA is a decidable language

Hint: Let C be a DFA that accepts strings


that is in L(A) but not in L(B), and also
strings that is in L(B) but not in L(A).
Then, L(C) ={ } if and only if L(A) = L(B)
Equality Test for DFA (2)
Proof: Based on the hint, we give a TM
DEQDFA that decides EQDFA as follows:
DEQDFA = “
On input  A,B
1. Construct C (how?)
2. Run DEDFA (Emptiness-Tester for DFA) on 
C
3. If DEDFA accepts, accept. Else, reject”
Acceptance by CFG
Let ACFG be the language
{G, w| G is a CFG that generates w }

Theorem 6: ACFG is a decidable language

Hint: We need to avoid testing infinite


derivations… If G is in Chomsky normal
form, any derivation of w takes exactly
2|w| - 1 derivation steps
Acceptance by CFG (2)

Proof: Based on the hint, we give a TM


DACFG that decides ACFG as follows:
DACFG = “
On input  G, w
1. Convert G into G’ = (V,T,R,S) in CNF
2. Generate all derivations of G’ with
2|w|-1 derivation steps (Note: # of such
derivations is less than (4|V||T|)2|w|-1. That is, a
finite number)
3. If any derivation generates w, accept.
Else, reject”
Emptiness Test for CFG
Let ECFG be the language
{G| G is a CFG and L(G) = { } }

Theorem 7: ECFG is a decidable language

Observation: Suppose that we can mark


all the variables in G that can generate a
string of terminals. Then, L(G) = { } if
the start variable is not marked
Emptiness Test for CFG (2)
Proof: We use similar idea as we test if a
graph G is connected. Precisely, we give
a TM DECFG that decides ECFG as follows:
DECFG = “On input G
1. Mark all terminals of G
2. Repeat until no new variable is marked
•Mark variable A if G has a rule
A  U1U2…Uk and all Ui’s are marked
3. If the start variable is not marked,
accept. Else, reject”
Equality Test for CFG?
Let EQCFG be the language
{A,B| A and B are CFGs and L(A) = L(B) }

Is EQCFG is a decidable language?

Unfortunately, no…
Note that we cannot apply similar trick as we
prove EQDFA is decidable
We shall show EQCFG is undecidable later…
TM can simulate CFG
•Previously (a long time ago), we have
shown that given a DFA, we can always
find a CFG that decides the same
language

•How about, if we are given a CFG, can we


find a TM that decides the same language?
–The answer is YES!
TM can simulate CFG (2)
Theorem 8: Given a CFG G, we can
construct a TM that decides the same
language. In other words, every CFL is
a decidable language
TM can simulate CFG (3)
Proof: We find a TM MG with  Gstored in
it initially; MG then performs as follows:
MG = “ On input  w
1. Run DACFG (Accept-by-CFG checker) on 
G, w
2. If DACFG accepts  G, w
, accept.
Else, reject ”
We can see that MG decides the same
language as G. This completes the proof
Language Hierarchy (revisited)
Set of Languages (= set of “
set of strings”
)

Set of Decidable Set of Recognizable


Language Language
{0n1n2n}
{ww}
{w with even |w|} ???

??

{0x1y}
{0n1n}
Set of Regular Set of Context-
Language {w | w = wR} Free Language
Next Time
•Undecidable Languages
–Languages that CANNOT be decided by
ANY Turing Machine
–Example 1: Turing-recognizable, but not
Turing-decidable
–Example 2: Non-Turing recognizable
(that is, even more difficult!!)

You might also like