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

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

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

a spike. Similarly, given an SNP system Π that performs sequential routing of a<br />

spike, there exists a Petri net N Π simulating Π that performs sequential routing<br />

of a token.<br />

Proof. (An illustrati<strong>on</strong> of the proof can be seen in Fig. 3.) Given a Petri net N<br />

with places p, q and a transiti<strong>on</strong> t, we have p ∈ •t and t ∈ •q. Given a marking<br />

M(p) over p, N can be simulated by an SNP system Π N having neur<strong>on</strong>s σ x , σ y<br />

where R x = {a + /a → a}, α x = M(p), and (x, y) ∈ syn such that t is fired<br />

iff σ x applies rule a + /a → a. M(p) serves as the number of spikes in σ x . Rule<br />

R x in σ x is of the form a + /a → a: R x c<strong>on</strong>sumes <strong>on</strong>e spike whenever α x ≥ 1,<br />

and produces <strong>on</strong>e spike. Note that variable overloading is performed because of<br />

the use of R x to mean the set of rules in σ x and to mean the <strong>on</strong>ly rule in σ x .<br />

Transiti<strong>on</strong> t is fired if there is at least <strong>on</strong>e token in p. The firing of transiti<strong>on</strong> t<br />

sends <strong>on</strong>e token to output place q. Similarly, rule R x is applied if neur<strong>on</strong> σ t has<br />

at least <strong>on</strong>e spike. The neur<strong>on</strong> sends a spike to its output neur<strong>on</strong> σ y after R x is<br />

applied. For N, if M 0 = (1, 0) (i.e. <strong>on</strong>ly p is marked) and the final c<strong>on</strong>figurati<strong>on</strong><br />

is (0, 1), Π N similarly has C 0 = 〈1, 0〉 and a final c<strong>on</strong>figurati<strong>on</strong> of 〈0, 1〉.<br />

The reverse can be shown in a similar manner i.e. given an SNP system Π<br />

that routes a spike sequentially, there exists a Petri net N Π that can simulate<br />

Π: a forgetting rule is a transiti<strong>on</strong> with an outgoing arc weight of zero so no<br />

token is ever produced, and the envir<strong>on</strong>ment is a sink place. For spiking rules,<br />

the regular expressi<strong>on</strong> E is an additi<strong>on</strong>al c<strong>on</strong>diti<strong>on</strong> before a transiti<strong>on</strong> t in a<br />

Petri net is fired: if place p ∈ •t, then t is enabled iff a M(p) ∈ L(E) i.e. when<br />

rule R x is applied then transiti<strong>on</strong> t is also fired.<br />

⊓⊔<br />

Fig. 3. The “basic” transformati<strong>on</strong> idea from a Petri net performing sequential<br />

routing to an SNP system (and back).<br />

Lemma 2. Given Petri net N that performs AND-split (AND-join) routing of<br />

a token, there exists an SNP system Π N simulating N that performs AND-split<br />

(AND-join) routing of a spike.<br />

Proof. (An illustrati<strong>on</strong> of the proof can be seen in Fig. 4.) The proof follows<br />

from Lemma 1 and the following c<strong>on</strong>structi<strong>on</strong>s: Given an AND-split Petri net N<br />

with places i, j, k, transiti<strong>on</strong> t, such that i ∈ •t and j, k ∈ t•, the AND-split SNP<br />

system Π N that simulates N has neur<strong>on</strong>s σ x , σ y , σ z where R x = {a + /a → a},<br />

with (t, j), (t, k) ∈ syn. For N we have M 0 = (1, 0, 0) i.e. <strong>on</strong>ly i is marked, with<br />

a final marking of (0, 1, 1) after the firing of t. N performs an AND-split, sending<br />

<strong>on</strong>e token each to j and k. For Π N we have C 0 = 〈1, 0, 0〉 and the firing of σ t<br />

sends <strong>on</strong>e spike each to σ j and σ k . The final c<strong>on</strong>figurati<strong>on</strong> is 〈0, 1, 1〉, thus Π N<br />

performs an AND-split.<br />

150

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

Saved successfully!

Ooh no, something went wrong!