09.09.2014 Views

13th International Conference on Membrane Computing - MTA Sztaki

13th International Conference on Membrane Computing - MTA Sztaki

13th International Conference on Membrane Computing - MTA Sztaki

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.

A new approach for solving SAT by P systems with active membranes<br />

Clearly, the main issue with Π(n) is that it can not be c<strong>on</strong>structed in polynomial<br />

time in n. As we have menti<strong>on</strong>ed, the reas<strong>on</strong> of this is that Π(n) creates<br />

complete clauses from the clauses of the input formula, and the number of these<br />

complete clauses can be exp<strong>on</strong>ential in n. On the other hand, <strong>on</strong>e can note that<br />

the answer of Π(n) depends <strong>on</strong>ly <strong>on</strong> whether or not every membrane with label<br />

n + 3 c<strong>on</strong>tains at least <strong>on</strong>e object regardless of whether the set of these objects<br />

c<strong>on</strong>tains every complete clause or not. Thus, <strong>on</strong>e way to turn Π(n) into<br />

a polynomially (semi-)uniform soluti<strong>on</strong> of SAT might be to modify Π(n) such<br />

that it reuses somehow the original clauses of the input formula in every step of<br />

the computati<strong>on</strong>. We dem<strong>on</strong>strate our idea that might be such a soluti<strong>on</strong> using<br />

the P system Π(3) from Example 1. Let Π ′ (3) be a slight modificati<strong>on</strong> of Π(3)<br />

working as follows. Assume that Π ′ (3) is started with the same input symbols<br />

xyz, ¯x, ȳ, ¯z as in Example 1. In the first step, Π ′ (3) does not create new clauses<br />

from ȳ and ¯z, it <strong>on</strong>ly makes two copies of them, <strong>on</strong>e marked with a prime, and<br />

another <strong>on</strong>e marked with a double prime, i.e., Π(3) creates ȳ ′ and ȳ ′′ from ȳ,<br />

and ¯z ′ and ¯z ′′ from ¯z. Then Π ′ (3) separates the new copies into the two new<br />

membranes according to that whether they are marked with a prime or with a<br />

double prime. The remaining clauses are separated in the same way as it is d<strong>on</strong>e<br />

in Π(3). Thus, after the first step the membrane with label 3 looks as follows:<br />

[[xyz, ȳ ′ , ¯z ′ ] 4 , [¯x, ȳ ′′ , ¯z ′′ ] 4 ] 3 . Then, in the next step, Π ′ (3) creates the clauses ¯z ′<br />

and ¯z ′′ from ¯z ′ , ¯x ′ and ¯x ′′ from ¯x, ¯z ′ and ¯z ′′ from ¯z ′′ (of course, here Π ′ (3)<br />

requires such rules also that can rewrite ¯z ′ and ¯z ′′ similarly as ¯z was rewritten<br />

before). Moreover, both ȳ ′ and ȳ ′′ are rewritten to ȳ by two corresp<strong>on</strong>ding rules.<br />

Then the symbols are separated into the new membranes as follows:<br />

[[[xyz, ¯z ′ ] 5 , [ȳ, ¯z ′′ ] 5 ] 4 , [[¯x ′ , ¯z ′ ] 5 , [¯x ′′ , ȳ, ¯z ′′ ] 5 ] 4 ] 3 .<br />

Finally, after the third step of Π ′ (3), the membrane with label 3 looks as follows:<br />

[[[[xyz] 6 , [¯z] 6 ] 5 , [[ȳ ′ ] 6 , [ȳ ′′ , ¯z] 6 ] 5 ] 4 , [[[¯x ′ ] 6 , [¯x ′′ , ¯z] 6 ] 5 , [[¯x ′ , ȳ ′ ] 6 , [¯x ′′ , ȳ ′′ , ¯z] 6 ] 5 ] 4 ] 3 .<br />

Now, since every membrane with label 6 c<strong>on</strong>tains at least <strong>on</strong>e object, Π ′ (3) can<br />

c<strong>on</strong>tinue the computati<strong>on</strong> and send out the symbol no in the same way as Π(3)<br />

does it.<br />

If we generalise the above idea to Π ′ (n) (n ∈ N), <strong>on</strong>e can see that the<br />

number of objects used in Π ′ (n) is linear in n + m, where m is the number of<br />

clauses of the input formula. Thus, it seems that we can c<strong>on</strong>struct a polynomially<br />

semi-uniform family of P systems with active membranes that is based <strong>on</strong> the<br />

P systems presented in this paper and solves SAT in linear time in n. Our<br />

future plan is the exact definiti<strong>on</strong> of this family of P systems and showing its<br />

correctness. Moreover, we are planning to implement these P systems <strong>on</strong> certain<br />

systems using parallel hardware since we would like to see whether our new<br />

approach can be utilized in practice as well.<br />

Acknowledgements. The authors are grateful to the reviewers for their valuable<br />

comments that improved the manuscript.<br />

219

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

Saved successfully!

Ooh no, something went wrong!