27.03.2014 Views

SEKE 2012 Proceedings - Knowledge Systems Institute

SEKE 2012 Proceedings - Knowledge Systems Institute

SEKE 2012 Proceedings - Knowledge Systems Institute

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

satisfies R( D) R( C) S 1. Compared to C, D is obtained<br />

1<br />

by deleting one column vector from C. Assume the deleted<br />

column vector is the j th column of C and denote it by C .<br />

Next we will construct a non-trivial integer solution of<br />

<br />

A Y 0 with D, C and j defined above.<br />

S1 S1<br />

j<br />

<br />

Lemma 4 For the equation A Y 0<br />

S1 S1<br />

, denote the k<br />

th<br />

element<br />

of<br />

YS 1<br />

by<br />

y<br />

S 1 : k<br />

. Let<br />

y<br />

S1 :<br />

j<br />

det( D)<br />

. And for i <br />

1,<br />

S i j<br />

1 ,<br />

let y det( D ( C )) , where<br />

S1 : i i j<br />

det( D ) is the determinant of<br />

D, D ( C ) denote the resulted matrix after replace the i th<br />

i<br />

j<br />

<br />

column of D with vector C j . Then Y y , y ,..., y<br />

<br />

<br />

is a non-trivial integer solution of A Y 0<br />

S1 S1<br />

.<br />

C<br />

k<br />

S S1:1 S1:2 S1:<br />

S<br />

1 1<br />

Proof : Denote the k th column vector of the matrix C by<br />

. Denote th e resulted vector after deleting the j th element<br />

from YS 1<br />

by Y . According to th e relationship between D<br />

S 1<br />

/ j<br />

and C, CY 0<br />

S1<br />

has the following equivalent transformations:<br />

y<br />

S1:1<br />

<br />

<br />

<br />

S1<br />

y<br />

S1:2<br />

<br />

<br />

CY 0 C , C ,..., C y C<br />

0<br />

S1 <br />

1 2 S<br />

S 1 <br />

1:<br />

i i<br />

... <br />

i1<br />

y<br />

<br />

<br />

S1:<br />

S1<br />

<br />

<br />

<br />

y C y C 0 DY y C<br />

0<br />

<br />

1, <br />

i j j S<br />

1,..., 1,...| | 1<br />

DY y C<br />

/ :<br />

S j S j j<br />

1<br />

1<br />

: i : / :<br />

S i S j j S j S j j<br />

1 1 1<br />

1<br />

Let y det( D)<br />

, then the above equation becomes<br />

S1 : j<br />

DY det( D)*<br />

C . Because D is a full-rank matrix, the<br />

S1 / j j<br />

solution of DY det( D)*<br />

C is unique. According to the<br />

S1 / j j<br />

Cramer’s rule [10] , the i th element in the solution is as follows:<br />

det( ( det( )* ))<br />

D D C<br />

1 : i<br />

det( D )<br />

i<br />

j<br />

y det( D ( C )) <br />

i j<br />

S<br />

i 1, 2,..., j 1, j 1,... S .<br />

where<br />

1<br />

Thus, y det( D ( C ))<br />

S1 : i i j i 1, 2,..., j 1, j 1,... S1<br />

<br />

constitutes a solution of DY y C<br />

S1 / j S1:<br />

j j<br />

when<br />

y det( D)<br />

. Because DY y C is equivalently<br />

S1 : j<br />

S1 / j S1:<br />

j j<br />

<br />

transformed from CY 0 , so y det( D ( C ))<br />

S1<br />

S1 : i i j<br />

i 1, 2,..., j 1, j 1,... S1<br />

and y det( D)<br />

constitute one<br />

S1 : j<br />

<br />

solution of CY 0 (i.e., Y <br />

S1<br />

S<br />

is a solu tion of CY 0<br />

1<br />

S1<br />

). In<br />

<br />

<br />

addition, because A Y 0 and CY 0 are equivalent, so<br />

S1 S1<br />

S1<br />

Y <br />

S 1<br />

is also a solution of A Y 0<br />

S1 S1<br />

.<br />

Furthermore, according to th e construction method of D<br />

and C , their elements come from . According to<br />

j<br />

A<br />

S 1<br />

j<br />

<br />

T<br />

Definition 1, the elements of A<br />

S 1<br />

are all in tegers, so D and<br />

D ( C ) are integer matrixes. Hence y det( D)<br />

and<br />

i j<br />

S1 : j<br />

y det( D ( C )) are all in tegers. Finally, y det( D)<br />

is<br />

S1 : i i j<br />

S1 : j<br />

nonzero because D is full-ranked. Therefore, Y is a nontrivial<br />

integer solution of A Y 0<br />

S1<br />

<br />

S1 S1<br />

. ■<br />

According to T heorem 4, the solu tion Y constructed in<br />

S1<br />

Lemma 4 must be positive or negative integer vector. If Y <br />

is<br />

S1<br />

positive let Y Y<br />

<br />

, else let Y 1<br />

Y<br />

<br />

. Then Y must be a<br />

S1 S1<br />

S 1 S 1<br />

S<br />

1<br />

positive integer solution of A Y 0<br />

S1 S1<br />

. Each element of Y <br />

S1<br />

corresponds to a place in S 1<br />

. For each place in S S , we add<br />

1<br />

a zero elem ent to Y and denote th e resulted vector by Y .<br />

S1<br />

Then it’s obviously that Y is an S-invariants supported by S .<br />

1<br />

As a result, we can get the following algorithm.<br />

Algorithm 2 Computation of a minimal-supported S-invariant<br />

INPUT: a minimal support of S-invariants S and its generated sub-matrix<br />

1<br />

OUTPUT: one integer S-invariant supported by S<br />

1<br />

PROCEDURES:<br />

Step1: Get the great linearly independent group of<br />

A<br />

S 1<br />

A's<br />

S 1<br />

row spaces with<br />

elementary column transformation method [10] . Denote the matrix composed of<br />

these row vectors by C;<br />

Step2: Get the great linearly independent group of C’s column spaces<br />

with elementary row transformation method [10] . Denote the matrix composed<br />

of these column vectors by D;<br />

Step3: In fact, D is obtained by deleting one column vector from C.<br />

Assume the deleted column vector is the j th column of C and denote it by C ;<br />

j<br />

Step4: For equation A Y<br />

S1 <br />

0 , denote the k<br />

th<br />

element of YS 1<br />

by y . Let<br />

S 1 : S1<br />

y det( D)<br />

. And for i 1,<br />

S i j<br />

S1 : j<br />

<br />

<br />

1 , let y det( D ( C ))<br />

S1 : i i j<br />

.<br />

<br />

T<br />

Construct the vector Y y , y ,..., y ;<br />

S1 <br />

S1:1 S1:2 S1:<br />

S1<br />

<br />

Step5: If Y <br />

S<br />

is positive, let Y<br />

S<br />

Y<br />

S<br />

,else let Y S<br />

1Y<br />

<br />

S<br />

;<br />

1<br />

1 1<br />

1 1<br />

Step6: Denote the element of Y S 1<br />

corresponding to place s S<br />

i 1<br />

by<br />

y ( s ) .Construct an |S|-dimensional column vector Y as follows:<br />

i S1<br />

y<br />

( s ) if s S<br />

S1 j j<br />

y<br />

( s ) <br />

j <br />

<br />

0 if s S<br />

j 1<br />

1<br />

where y ( s j<br />

) is the element of Y which corresponds to s S ;<br />

j<br />

Step7: Output that Y is an S-invariants supported by S .<br />

1<br />

Considering the Petri net and S s , s , s , s , s<br />

2 3 4 5 6 7<br />

in Fig.2.<br />

In Step1, the great linearly independent group of A<br />

S 2<br />

's row<br />

spaces constitutes C as follows.<br />

1 1 0 0 0 1 1 0 0 0 <br />

<br />

0 1 1 1 0 0 1 1 1 0<br />

C D C <br />

5<br />

0 0 1 1 0 0 0 1 1 0 <br />

<br />

0 0 0 1 1 0 0 0 1 1<br />

In Step2, the great linearly independent group of C’s<br />

column spaces constitute D shown above. In Step3, it’s easy to<br />

344

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!