Minimization of DFA

Indistinguishablecity relation
Definition : 27
Let the
any languageasin The
indistinguishable'ty
relation In on E* is defined follows :

*
for all my c- 8 , we have a Ivy if and only if nandy
are
indistinguishable with
respect to L.

( for all 2- C- S*, either both sit and ytz c- L or neither of

them
belongs to 1.)

consider the
relation .
equivalence classes
of the
indistinguishable'ty
that there is an isomorphism between the
observe
•

equivalence classes
of It and the states of the minimum
state DFA A- !
 Checking equivalence of states

DFA A- = IQ, Sir go , F) we want to fund the

given a , .

minimum state DFA A ' such that HAD =L (A) .

called
Two states p ,
q are
distinguishable if there
•

exists a c- 3.* such that

( p a) c- F and 8*19 a) ¢ F.
*
s , ,

An
algorithm
Procedure mark
for
identifying all
distinguishable states :

Remove all states thatare inaccessible them 90 .

if PER and q 4-
For distinguishable
•
base case : p, q are

vice versa

a-
q→aq
'
a c- E
'
•
Iterative step Ifp p and for some
'
'
and P and g have already been shown to be

distinguishable ,
then mark p , q as
distinguishable .

Repeat me iterative
step until no previously unmarked pairs
are marked .
 Example of minimization of DFA
o
ai
90 91 92
o I 093
0 1
I 0

0 A t
1 I 0
96 9s 96 97
o
o

Procedure mask

I
words
✗
Distinguishing
£

9s ✗
96 ✗
95 ✗

qÉ
%

÷s%
 Example of minimization of DFA
0
At
90 91 92
o I 093
0 1
I 0

0 A t
' I °
ga qs q, qq
o
o

Procedure mask

1-:[
words
Distinguishing
e

↳ XXX 0
96 ✗ ✗
95 XXX x

% ✗ ✗ ✗

É
 Example of minimization of DFA
o
ai
90 91 92
o I 093
0 1
I 0

0 A t
' I °
ga qs q, qq
o
o

Procedure mask
TX

①
words
92 X X
Distinguishing
E

↳ ✗ ✗ ✗ 0

* × .
95 XXX x

%×
Ex
É
 Example of minimization of DFA
o
ai
90 91 92
o I 093
0 1
I
0 i
' I °
ga qs q, qq
o
o

Procedure mask
TX

①
words
92 X X
Distinguishing
E

↳ ✗ ✗ ✗ 0

* × .
95 XX X X 01
96 XXX XXX
¥x
É
 Example of minimization of DFA
o
ai
90 91 92
o I 093 .

0 1
I
0 i
' I °
ga qs q, qq
o
o

Procedure mask
TX

①
words
92 X X
Distinguishing
E

% XX X 0

* × .
95 XX X X 01
go ✗ ✗ X XXX
É
90=-94 95=95 95=47
É
 Minimization of DFA
Theorem : The procedure mark , applied to DRA A- CAR S go F)
any
=
, , ,

terminates and determines all

pairs of distinguishable states .

Proof : The iterations of the procedure terminates if in an

iteration , no new
pair is marked in an iteration Since there
.

are
finitely many pairs the of states , the procedure mark terminates .

C. In every iteration 1 apart from last one) at least one pair is marked ) .

we have to prove that if

then it is marked by the procedure mark
pair of states is
a

Completeness
.
→
distinguishable ,

If fp, g) are distinguishable , then

smallest possible length that
there exists a
and
string n
of
distinguishes p q
to show that the pair CP, g) is
.

We on the

markedargue If n=E
ieugth of
then ,
n

Base care :O all entries

distinguishedis bytruec-far marked
•
are .

Induction :
suppose the hypothesis lnkm .

Nowfor t.nl F- ay and let Iyl = m -1

=m, let .

Now suppose ftp. a) =t and 8 ( q a) u Now t and u are ,

=
.

the pair Lt, a) is marked

distinguished by and y by hypothesis .

When
fp g)
, visited
is in the next iteration
,
then the pair will be marked .
Sounders Procedure mask marks pairs of states
only those
that are
distinguishable .

mark
Proof :
Define an iteration i
of the
, algorithm to be one in the

following way .

-
iteration i corresponds to
-

iteration
distinguishing
0 in the initialization
strings of length i.
C. Marks states that
are
distinguishable by E.)
need to check with all letters of E to
we
only
•

identify distinguishable
previously marked states
states with respect to the
.

The
procedure mark marks states to be
a pair of
-

distinguishable ini iteration i only if there enistn a

string of length distinguishes them .

Hence procedure mark is sound .

 Minimization of DFA

Procedure reduce :

given a Di-AM-CQ.GS , go F) we construct

,
a reduced DFA
"
M =
lot,S 8,1901, F) as follows :
,

i. Call procedure mark to obtain the set of equivalence classes .

a' =
{ [qifq is accessible from Go }
'
F -41%1 Iq is on F }
s God a) =&(q a) ]
'

, ,

↳ This is
consistent with the
definition of
indistinguishablety since if q Iip ,
then 819 a) Iit ( Psa)
, .
 Minimization of DFA

'
Theorem : The DFA M constructed by the procedure reduce
with inaccessible states removed , is the minimum state
DFA for its language .

Proof :
First we show that Llm D= LCM ) .

• we show that 8*11%2,4=[8*19003] big conduction on Kol .

From the above , it follows that HMD =L ( M) .

•
Minimality follows from a previous theorem .

Complete the
proof ( Homework )
-
 Example of minimization of DFA
o
ai
90 91 92
o I 093
0 1
I 0

0 A 1
1 I 0
96 9s 96 97
o
o

90=-94 93=-95 9,797

o
ai
1
904 0 917 92
O
l
0
GO
0 935 •
96
 strong bisimilaiuty
•
Communication potential of intermediate states
•
states reached are equivalent
set actions
same
of
-

cannot be done
by trace equivalence
•

q*t→
↳
.
+7
*
* to f- → o
 strong kisimulatiou

relation R the set states automata

A
'
binary
kisiwmlatiou
over
whenever
of of
Rs 2 and ✗ is
an

action:
m a riff s
,
an

if S, →✗s
,
'
,
then there is a transition 82^-521 such that si RSL;
if sn✗→si ,
then here is a transition s
,
-4s '

,
such that si R si .

S , C- F iff S2 c- F.
' '
Two states S and S are kisimitar written
,
Sns
iff there is a
,
Kisi mutation that relates them .

is the disconnection relation also called the

largestrelation
~

ki
similarity .

~
= U R is akisiniulntion relation
R
 Example

¥ → t

a
a a

L s v t
,
>

si
b Ob ⑨ bb
Sz

To show Smt
,
we need to define a
strong kisiwmlation
relation R={ Is,t) Csi ,tD , ,
Csz,tD } .
 strong kisiwilarity
Theorem : The relation ~

1. is an equivalence relation
2. is the
largest strong bisi nutation
2. satisfies the
following property :

Sirs, iff ,
for each action ✗
,

if Sissi ,
then there is a transition Sissi
such inert sin si ;

if Sz ASI ,
then there is a transition Sissi
such that si ~
SE .

S , c- F iff Saff .
 strong kisiiuilarity
Theorem : The relation ~

1. is an equivalence relation
2. is the
largest strong Kisi nutation
2. satisfies the
following property :

Sirs, iff ,
for each action ✗
,

if Sissi , then there is a transition Sissi

such inert sin si ;

if Sz ASI ,
then there is a transition sites !
such that si
S , c- F iff SZEF .
n
si .

Proof : I =
{ Is s) , Is c- Q } is a biaiaiulation relation
~ is the union
of
all biaiuuelation relations
let 1s , .si ER
ksa.si/ Csi ER}
} Symmetric
'
consider pi = , .

'
A- is indeed a késiwmlaliou .
 strong kisimilarity
Theorem : The relation ~

1. is an equivalence relation
2. is the
largest strong bisi nutation
2. satisfies the
following property :

Sirs, iff ,
for each action ✗
,

if Sids si , then there is a transition Sissi

such inert sin si ;

if Sz ASI ,
then there is a transition s
,
-4s !
LEA such that si si
S c-
, Fiff .
n
.

Proof : I -_
{ ls.SI/sc-Q1isakiaimulation relation

~ is the union
of
all biaiwmlation relations
let lsi.si ER
consider
A-
'
is
R-i-ksa.si/Csi,sDER}
indeed a késiniulaliou .
.

} Symmetric
I =
{ Csi ;) /,
s Csi .si ) C- Rand Kiri ) c- R! for soonest } ,

kiaiuuelatéous
'
where R and R are .

It follows that I is a his imitation .

 strong kisimilarity
Theorem : The relation ~

1. is an equivalence relation
2. is the
largest strong bisi nutation
2. satisfies the
following property :

Sirs, iff ,
for each action ✗
,

-4 si then
if S,
,
there is a transition Sissi
such inert sin si ;

if Sz ASI then there

,
is a transition sits si
s c- f- iff sett such that si n
si .
,

Proof :o) ~ = U { KIR is a bis invitation }

This shows that each his invitation is included in ?

We need to show that the RHS is a biaiwmlation .

↳
We show that if Csi E) Em and
,
s
, si ,
then

Fsd : Sst si and Csi si ) c-

,
~
.

If SC.si c- ~
,
then & E) ER for
,
some kisimulation relation R
and thus @ F. sit c-R and hence Csi SITE ,
?
 string kisiwitarity
over that set of states of automaton
A
binary relation an
n' a
string his invitation iff whenever s
, Rsz and n c- E. *,

if s
,
-7 si ,
then there is a transition Sissi such that SIRSI;
if Sisi ,
then there is a transition Sissi much that sik si;
s
, iff set F.
c- F

states s and dare taisiiuilas iff there

Two
string
his invitation that relates them
exists a

string ,

Theorem : Two states s and dare string kisiuiiiar

iff they bisinielar
are
strong .
 strong his invitation and
language equivalence
also define kisiiuulation between
we can
strong states of different
automata over the same
alphabet .

We
say two automata A and B over
the same
alphabet to be strongly
biainrilar iff their initial states are
strongly bisinrilar .

Theorem
automata kisimilar , then they are
If two are
strongly also
language
equivalent . However ,
the converse does not hold true .

We first prove the following lemma .

Lemma : If Sirsa then for all n

si) ER
,

si →
si sissi and Csi ,
tn
,
: SIS
,
,
7s ; st .

nisi : St ez Asi ,
Is ,
'
s .
t .
s ,
Assi ,
and ( Si Si ) C- R
,

where R is a bisimutation .

much that s , -9s,

'
Then consider an n and si c- f.