Non-Deterministic Finite Automata (NFA)

Non-deterministic Finite Automata (NFA)
Models of Computation
NFA versus DFA

In a DFA, at every state , for every symbol , there is a unique -transition
This is not necessarily so in an NFA. At any state, an NFA may have multiple -transitions, or none.
We may think of the non-determinism as a kind of parallel computation wherein several processes can be running concurrently. When the NFA splits to follow several choices, that corresponds to a process forking into several children, each proceeding separately. If at least one of these accepts, then the entire computation accepts.

are labelled by symbols from
. I.e. an NFA may have -transitions.
In a DFA, transition arrows are labelled by symbols from
i.e. there is a unique
such that
; in an NFA, they
Example: All strings containing a 1 in third position from the end

0 ' ) % # ! &$" 1 2) ( (
NFAs are more compact - they generally require fewer states to recognize a language.
Models of Computation 3
Example:
is a multiple of 2 or 3
Using -transitions and non-determinism, a language dened by an NFA can be easier to understand.
888 77B4
588 97B4
8 58 69B4
88 5 7A@4
8 55 6764
58 5 9664
558 79B4
555 7764
DFA:
0 0
4
1 3) 0
1 2) 0
0 ) )
NFA:
Denition: NFA
Some denitions and notations
(not necessarily all distinct) such that
hc&D u2 x " 3B u x yr
c e e hdD x rre
cb dD
For
where each
c h&D
c c h&D u h&D u u3 x 62B6rrrI u 6 I u x I e e e I u u x e e e u D 3y6rrwvtsa
-transitions in
from
to
iff there are
C D qpi
iff
or there is a sequence
h rre e e
cg &D e f
where
is a nal state. Here
is dened by: of one or more
, the language accepted by
W V I T I S I I GE XAUB66RQPHFD
Fix an NFA
. , consists of all strings over satisfying
W I E S pnm y6y9y& q7D x
i$w
For
we dene
G HE
presented as a relation, i.e. a subset of
GfvWWiuhgf E G
G tlk
r se
Note:
r pnm W G w qoD XHE j
G lk
hdD c
W C `9E
cb T dD B
(v)
G e yXS
(iv)
is the set of nal states. is the power set of . Equivalently can be
W G XHE j
W E ighgf
G T iB
(iii)
(ii)
(i)
is a nite set of states is a nite alphabet is the start state is the transition function
W V I T I S I I G dAU6AQHE
A nondeterministic nite automaton (NFA) is a 5-tuple
where
before or after each one.
Equivalence of NFAs and DFAs: The Subset Construction
Observation. Every DFA is an NFA! Say two automata are equivalent if they accept the same language. Theorem(Determinization). Every NFA has an equivalent DFA. Proof. Fix an NFA DFA , we construct an equivalent
in
iff
in
D W C È
such that
W uC j E
c e e e e e dD rrrrre B
c E &D 9$Xe D
W z AI z 6I z AI z QI z HFD V S GE W ~A{6}6X|QX{HFD C z V I z I z S I z I z GE
D z V e ppt~3z ~ qoD m G pn
c dD u
z G k e pnm } q7D z G
c dD r
hdD c
e pn z 2X qoD m
C j
z pn | qoD m
x 6BrrrI u I e e e
exist
such that
cb dD
u x e e e u D 3yArrwpa
Exercise. Writing
cb dD
that
(thus, by abuse of notation, we write , we have
).
is equivalent to: there
qpDWaI9 S E $a
G d
Note: In case
is a DFA, for any
and
, there is a unique
$a
cb e G &D 2F ~9 S D W a I E
We shall sometimes write
, for
symbols in
occur in the correct order, but with 0 or more -transitions
There is a sequence of transitions from
cb &D
Intuitively
means: to in in which the
. such
z
z V
z
C j
C j
z V ~ x
c e e e e e &D rrBrre
. Suppose
c &D u
c &D r
z
sequence of
W uC j E Y
z 3e D z WÈ C Y k W C vÈ Y
1 3)
5 B64 ) 0 ) 1 8 5 @r4 64 ) 0 8 B4 ( 5 64 C ) 0 C j
DFA
8 @r4 0
NFA
Note. State
T Urq
Suppose
Proof of
Example. All words that begin with a string of 0s followed by a string of 1s.
is redundant.
. Then
for some
. Hence
z &
WuCE Y WuC j E Y z V ~&
z V{ h&D c
c h&D z
and so,
Now take any non-null
x rre u p e e D
Since
is deterministic, feeding
x I e e e I u y6rrr
-transitions
e e e e e rrBrre
z C j
where
-transitions
i.e.
to it results in the sequence of
in
W `C
in
W uC
D D
& u
c E u &D 9$Xe c E &D 9}z X& e
By denition of
. . .
, from (1), we have . . .
, and so
e e rre $
,
u $ u
z S
. Then there is a
, and so , (2) (1)

9
10
Proof of Suppose
equivalently, for some
. Hence
Now suppose some non-null of
. I.e. there is a sequence containing some
-transitions of the form (2) with . Now since with
i.e. with
, by denition of in
, there is some
. Working backwards in this way, we can
build up a sequence of
-transitions like (1), until we deduce that -transitions with
Thus we get a sequence of
, and hence
Closure under regular operations revisited
Using nondeterminism makes some proofs much easier. Theorem. Regular languages are closed under union.
of (the respective state transition graphs of)
u C
start state of
and
and
, and a -transition from
to each
adding a new start state
to the disjoint union
W E C
W u E C
Dene
that accepts
u wC
Take NFAs
and
by
W`CE $ Y u &D c z
D w
h&D c
V x C z C f x dD u x C c x 2 x z S x i V x z W `C j E x Arre u p e e D Y WÈ C Y z e z 3d~V
. Then
i.e.
V x z B
V hc&D z z z z V W`C j E Y W`CE kWuC j E Y Y WC j E Y x x iB f f u x u x zV x C j u C
V x z i
, and hence
because
. Thus (2) shows that
, or
11
12
Theorem. Regular languages are closed under concatenation.
of
Theorem. Regular languages are closed under star. First attempt:
accepts
from each accepting state to the start state.
What is wrong with this?
accepting, and by adding a new -transition
u C
Obtain
u C W u V I u I u S I I u GE D u QA6r6RQHFwC
Take an NFA
that
. Construct
that accepts .
from
by making the start state
e e e rr$ u u u D u
C C
of
. The accepting states of
are those
u C
each accepting state of
to the start state
state of
, and by adding an -transition from
u C
by making the start state of
the start
be obtained from the disjoint union of
u C W E C
e W u E C
An NFA
that accepts
u C
Take NFAs
and
can and

13 14
( 8 (
( ( 0 )
E.g.
W u E C Y 9F D W E E 9FD W t9E
which is not in
( 0 )
The above construction gives the NFA
Consider the two-node two-edge NFA
u C
that accepts
vlUe
(namely the language
dened by the regular expression
But
accepts
The NFA
accepts
accepts
W u C AwE
Take an NFA
W u AI u 6I u 6RQI u hFD u C V S I GE
Dene
W U T RAI T 66RQ2 T lHE u V I S I I u G
where
Second (correct) attempt:
C C
Proof: Regular languages are closed under star
).
that accepts
and
u V
and
D
and
D
and
D
and
D
T D Bpi D T pi u V d u V d u d G u &
D W I E u69S u W I E u &69qS W I E u 69qS W I 69E u S
16 15
u {

Non-Deterministic Finite Automata (NFA)

Uploaded by

Copyright:

Available Formats

Non-Deterministic Finite Automata (NFA)

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Non-Deterministic Finite Automata (NFA)

Uploaded by

Copyright:

Available Formats

Non-deterministic Finite Automata (NFA)

NFA versus DFA

In a DFA, at every state , for every symbol , there is a unique -transition

are labelled by symbols from

. I.e. an NFA may have -transitions.

In a DFA, transition arrows are labelled by symbols from

i.e. there is a unique

Example: All strings containing a 1 in third position from the end

Some denitions and notations

(not necessarily all distinct) such that

hc&D u2 x " 3B u x yr

c c h&D u h&D u u3 x 62B6rrrI u 6 I u x I e e e I u u x e e e u D 3y6rrwvtsa

iff there are

is a nal state. Here

is dened by: of one or more

, the language accepted by

. , consists of all strings over satisfying

W I E S pnm y6y9y& q7D x

presented as a relation, i.e. a subset of

r pnm W G w qoD XHE j

is the set of nal states. is the power set of . Equivalently can be

A nondeterministic nite automaton (NFA) is a 5-tuple

before or after each one.

Equivalence of NFAs and DFAs: The Subset Construction

 D z V e ppt~3z  ~ qoD m G pn

 z G k e pnm } q7D z  G

(thus, by abuse of notation, we write , we have

is equivalent to: there

is a DFA, for any

We shall sometimes write

occur in the correct order, but with 0 or more -transitions

There is a sequence of transitions from

means: to in in which the

WuCE Y WuC j E  Y z V ~&

Now take any non-null

to it results in the sequence of

c E u &D 9$Xe  c E &D 9}z  X& e

, from (1), we have . . .

, and so , (2) (1)

equivalently, for some

Now suppose some non-null of

. I.e. there is a sequence containing some

-transitions of the form (2) with . Now since with

. Working backwards in this way, we can

-transitions like (1), until we deduce that -transitions with

Thus we get a sequence of

Closure under regular operations revisited

of (the respective state transition graphs of)

, and a -transition from

adding a new start state

to the disjoint union

V x C z C f x dD u x C c x 2 x z  S x i  V x z  W `C j E x Arre u p e e D Y W`E C  Y z e  z 3d~V

V hc&D z z z  z  V W`C j E  Y W`CE kWuC j E Y Y WC j E  Y x x iB f f u x u x zV x C j u C

. Thus (2) shows that

Theorem. Regular languages are closed under concatenation.

Theorem. Regular languages are closed under star. First attempt:

from each accepting state to the start state.

What is wrong with this?

accepting, and by adding a new -transition

hc&D u2 x " 3B u x yr

c c h&D u h&D u u3 x 62B6rrrI u 6 I u x I e e e I u u x e e e u D 3y6rrwvtsa

W I E S pnm y6y9y& q7D x

r pnm W G w qoD XHE j

D z V e ppt~3z ~ qoD m G pn

z G k e pnm } q7D z G

WuCE Y WuC j E Y z V ~&

c E u &D 9$Xe c E &D 9}z X& e

V x C z C f x dD u x C c x 2 x z S x i V x z W `C j E x Arre u p e e D Y W`E C Y z e z 3d~V

V hc&D z z z z V W`C j E Y W`CE kWuC j E Y Y WC j E Y x x iB f f u x u x zV x C j u C

W U T RAI T 66RQ2 T lHE u V I S I I u G

D W I E u69S u W I E u &69qS W I E u 69qS W I 69E u S