Professional Documents
Culture Documents
Abstract
In this paper we have defined input-connected fuzzy automaton and
established that in an input connected fuzzy automaton, element of an
orbit maps to an element of the orbit under the weak fuzzy automaton
endomorphism. Further, if an input x is unicore in the fuzzy automaton,
then the number of weak fuzzy automaton endomorphisms of A(x) is equal
to the number of states.
1 Introduction
In([2],[3]) Fleck has studied homomorphisms and isomorphisms of crisp au-
tomata especially for strongly connected automata. Many researchers have
worked on number of isomorphisms of crisp as well as fuzzy automaton. But
the problem of finding the number of homomorphisms of fuzzy automaton is
still unanswered. In this paper we make an attempt to find the number of weak
fuzzy automaton for fuzzy automaton with single input. We used the idea of
orbits and hands to find this number. We have defined input-connected fuzzy
automaton and discussed some properties such as an element of an orbit maps
to an element of the orbit under the weak fuzzy automaton endomorphism. The
number of weak fuzzy automaton endomorphisms of A(x) equals the number of
states of A, if x is an unicore input in A.
1
2 Preliminaries
Preliminary definitions and notations are presented in this section. Recall that,
ifA and B are two sets, then the fuzzy relation from A to B is a fuzzy subset
µ of A × B i.e.µ : A × B → [0, 1]. The number µ(a, b) denotes the degree to
which a is related to b.
2
N2 : IsF (A → B) denotes the set of all fuzzy automaton isomorphisms from
A to B and W IsF (A → B) denotes the set of all weak fuzzy automaton
isomorphisms from A to B
Definition 2.8. The x circle of q is the subautomaton Cx (q) = (Sxc (q), {x}, µ0 ),
where Sxc (q) = {t ∈ Sx (q) : µ(q, xk , t) > 0 and µ(q, xm , t) > 0, for some integer
m > k} and µ0 is the restriction of µ to Sxc (q) × {x} × Sxc (q) .
3
Note that for any q ∈ Q, µ(q) = φ or µ(q) = {q}. Let A(x) be a fuzzy
automaton. Then the core number of x, core (x), is the number of q ∈ Q such
that µ(q) = {q}. If core(x)= 1, then x is called an unicore in A.
Let C(x) be a component of A(x) and Co (x) be orbit in C(x). We call an orbit
Co (x) a self-loop, if |Co (x)| = 1.
y, 0.2
y, 0.9
q1 q4
y, 0.4
x, 0.4 x, 0.1 x, 0.3 x, 0.1
q2 q3
y, 0.6
Theorem 3.1. (i) If A(x) is connected, then A(x) contain only one orbit.
(ii) If A(x) is not connected, then each component of A(x) contain only one
orbit.
If p ∈ Q is a state in the component C(x) (or orbit Co (x)), then for the ease,
we shall write it as p ∈ C(x) (or p ∈ Co (x)).
Lemma 3.2. Let p ∈ Co (x) and µ(p, xk , q) > 0, for some k ∈ N and q ∈ Q.
Then q ∈ Co (x).
Proof. Suppose q 6∈ Co (x). Then there is a r ∈ Co (x) such that µ(r, x, q1 ) > 0
and µ(r, x, q2 ) > 0 with q1 6= q2 . This is impossible.
4
Theorem 3.3. Let A(x) be a connected fuzzy automaton and Co (x) be an
orbit in A(x). Then Co (x) is a self-loop if and only if there is a constant weak
fuzzy automaton homomorphism on A(x).
x, 0.1
q
x, 0.3
x, 0.1 x, 0.4
q1 q2 p
x, 0.4
Figure 2:
Let p, q ∈ Co (x), p 6= q and µ(p, x, q) > 0. Then µ(p, xk , p) > 0 with k > 1.
Thus µ(h(p), x, h(q)) > 0 and µ(h(p), xk , h(q)) > 0. i.e. µ(r, x, r) > 0 and
µ(r, xk , r) > 0, where h(p) = h(q) = r. By uniqueness of orbit in A(x), we
have p = q = r. Which is a contradiction. Conversely, suppose that Co (x) is
a self-loop and q ∈ Co (x). Then the map p ,→ q,∀p ∈ Q is the required weak
fuzzy automaton homomorphism on A(x).
Proof. Let k ∈ N such that µ(p, xk , p) > 0. Then µ(h(p), xk , h(p)) > 0. There-
fore, h(p) ∈ Co (x).
5
Clearly every x-connected fuzzy automaton is a connected fuzzy automaton,
but not conversely.
Definition 3.3. Let C(x) be a component of A(x) and q 6∈ Co (x). The hand,
Hx (q), of C(x) beginning at q with end HxE (q) ∈ Co (x) is a fuzzy automaton
Hx (q) = (Z, {x}, µ0 ), where µ0 = µ|Z×{x}×Z , Z = (Sx (q) − Co (x)) ∪ {HxE (q)}
and HxE (q) is a state p ∈ Co (x) such that µ(q, xk , p) > 0 with least nonnegative
integer k. In this case the value k is called the length of the hand Hx (q). We
shall write it as k = L(Hx (q)).
Definition 3.4. Let r ∈ Hx (q). Then the distance of r from Co (x) is defined
to be the least positive integer l such that µ(r, xl , HxE (q)) > 0. We denote it as
LHx (r) = l. We also take LHx (r) = 0, if r ∈ Co (x).
Example 3.2. Consider Q = {o0 , o1 , o2 , o3 , 04 , q11 , q12 , q21 , q22 , q31 , q32 , q33 , q42 , q43 , q44 },
Σ = {x} and µ is defined as follows.
6
q33
q22
x, 0.2
x, 0.1
q32
q21 x, 0.5
q31
x, 0.8
q44
x, 0.9
x, 0.3
o2
x, 0.3 x, 0.9
q43
x, 0.7
o1 o3
q42
x, 0.6 x, 0.5 x, 0.7
x, 0.5 x, 0.1
q12 q11 o0 o4
x, 0.1
Figure 3: x-connected
Example 3.3. Consider Q = {o10 , o11 , o20 , o21 , 022 , q111 , q112 , q211 , q212 , q221 , q222 , q223 , q121 ,
q122 , q123 }, Σ = {x} and µ is defined as follows.
7
q123
x, 0.4 q223
q122 x, 0.1
x, 0.7 q212 q222
x, 0.3
q121 x, 0.5
x, 0.6 q211 q221
x, 0.5
o11 x, 0.6
o21
x, 0.9 x, 0.7
x, 0.8 x, 0.6
x, 0.5 x, 0.1
q112 q111 o10 o20 o22
x, 0.9
There are 2 components C1 and C2 with Co1 = {o10 , o11 } and Co2 =
{o20 , o21 , o22 } For any hand Hx (q), if HxE (q) ∈ Co (x) then we say Hx (q) is
a hand corresponding to the orbit Co (x).
Proof. Let LHi (r) = mi , LHj (r) = mj and |Co (x)| = n. Then µ(r, xmi , HiE ) >
0. Therefore µ(s, xmi , h(HiE )) > 0 and hence due to Theorem 3.4, h(HiE ) ∈
Co (x). This is possible only if mi ≥ mj .
Proof. Let {q1 , q2 , . . . , qn } be the state set of Co (x) and {qn , qn+1 , . . . , qn+k } be
state set of H(qn+k ).
Let S = {h1 , h2 , . . . , hn+k }, where hi (qn+k ) = qi and h(q) = p, if µ(qn+k , xm , q)∧
8
µ(qi , xm , p) > 0. Clearly (by definition) S ⊆ W E F (A(x)).
Let h ∈ W E F (A(x)) and h(qn+k ) = qj . Let q ∈ Q with µ(qn+k , xm , q) > 0.
Then µ(qj , xm , h(q)) > 0. Also µ(qj , xm , hj (q)) > 0. Therefore h(q) = hj (q),
which implies that h = hj .
4 Conclusion
1. If A(x) is connected, then A(x) contain only one orbit.
References
[1] Bavel Z., Structure and Transition-Preserving Functions of Finite Au-
tomata,J. ACM 15(1),135-158, (Jan.1968).