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.

F.G.C. Cabarle, H.N. Adorna<br />

OR-split is synchr<strong>on</strong>ized by an AND-join, it is possible to have a deadlock. The<br />

free-choice property also avoids the possibility of deadlocks (see Fig. 2) and has<br />

been studied extensively in literature. See for example [5] and [21,19] to name a<br />

few. These nets are also known as extended free-choice nets in [15] and [4]. Next<br />

Fig. 2. (a) A net that is not well-handled. (b) A n<strong>on</strong>-free-choice net.<br />

we provide the definiti<strong>on</strong> of an SNP system, slightly modified from [18].<br />

Definiti<strong>on</strong> 4 (SNP system). An SNP system without delay of a finite degree<br />

m ≥ 1, is a c<strong>on</strong>struct of the form<br />

where:<br />

Π = (O, σ 1 , . . . , σ m , syn, out),<br />

1. O = {a} is the singlet<strong>on</strong> alphabet (a is called spike).<br />

2. σ 1 , . . . , σ m are neur<strong>on</strong>s of the form σ i = (α i , R i ), 1 ≤ i ≤ m,<br />

where:<br />

(a) α i ≥ 0 is the number of spikes in σ i<br />

(b) R i is a finite set of rules of the general form<br />

E/a c → a b<br />

where E is a regular expressi<strong>on</strong> over O, c ≥ 1, b ≥ 0, with c ≥ b.<br />

3. syn ⊆ {1, 2, . . . , m} × {1, 2, . . . , m}, (i, i) /∈ syn for 1 ≤ i ≤ m, are synapses<br />

between neur<strong>on</strong>s.<br />

4. out ∈ {1, 2, . . . , m} is the index of the output neur<strong>on</strong>.<br />

A spiking rule is a rule E/a c → a b where b ≥ 1. A forgetting rule is a rule<br />

where b = 0 is written as E/a c → λ. If L(E) = {a c } then spiking and forgetting<br />

rules are simply written as a c → a b and a c → λ, respectively. Applicati<strong>on</strong>s of<br />

rules are as follows: if neur<strong>on</strong> σ i c<strong>on</strong>tains k spikes, a k ∈ L(E) and k ≥ c, then the<br />

rule E/a c → a b ∈ R i is enabled and the rule can be fired or applied. If b ≥ 1, the<br />

applicati<strong>on</strong> of this rule removes c spikes from σ i , so that <strong>on</strong>ly k −c spikes remain<br />

in σ i . The neur<strong>on</strong> fires b number of spikes to every σ j such that (i, j) ∈ syn. If<br />

b = 0 then no spikes are produced. SNP systems assume a global clock, so the<br />

applicati<strong>on</strong> of rules and the sending of spikes by neur<strong>on</strong>s are all synchr<strong>on</strong>ized.<br />

The n<strong>on</strong>determinism in SNP systems occurs when, given two rules E 1 /a c1 → a b1<br />

and E 2 /a c2 → a b2 , it is possible to have L(E 1 ) ∩ L(E 2 ) ≠ ∅. In this situati<strong>on</strong>,<br />

148

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

Saved successfully!

Ooh no, something went wrong!