Outline

Some notes on Spiking Neural dP systems and Petri nets Francis George Cabarle, Henry Adorna Algorithms & Complexity Lab, Dept. of Computer Science, University of the Philippines Diliman [email protected], [email protected] http://aclab.dcs.upd.edu.ph

12 September 2011

Francis George Cabarle, Henry Adorna

Some notes on Spiking Neural dP systems and Petri nets

Outline

Outline

1

Introduction

2

Petri nets

3

SNP systems

4

Extending SNP systems & Petri nets

5

References

Francis George Cabarle, Henry Adorna

Some notes on Spiking Neural dP systems and Petri nets

Introduction Petri nets SNP systems Extending SNP systems & Petri nets References

Distributed computing

Great interest: break up & solve large problems. Combines multiple components, working together. Increase computational power (smaller execution time) or less work/component. Theory: graph theory, Petri nets et al. Practice: physically separated hard+software.

Francis George Cabarle, Henry Adorna

Some notes on Spiking Neural dP systems and Petri nets

Introduction Petri nets SNP systems Extending SNP systems & Petri nets References

Petri nets

Extensive theory research since 1939. Directed, bipartite, weighted graphs w/ 2 kinds of nodes: place & transition. Parallel, distributed, concurrent, syn/asynchronous, nondeterministic processes [Murata, 1989]. Used to model & study workflow processes, manufacturing & biological systems . . . [Murata, 1989][Piedrafita et al. 2010]

Francis George Cabarle, Henry Adorna

Some notes on Spiking Neural dP systems and Petri nets

Introduction Petri nets SNP systems Extending SNP systems & Petri nets References

Membrane computing

Introduced in 1998, aims to get computational ideas from ˘ the biological cell. [Paun, 1998] Computationally complete, parallel & nondeterministic models: P systems. Distributed problem solving: dP systems ˘ [Paun et al. 2010]. Focus on specific variant: Spiking Neural P systems (SNP systems) and SNdP systems [Ionescu et al. 2006][Ionescu et al. 2011]

Francis George Cabarle, Henry Adorna

Some notes on Spiking Neural dP systems and Petri nets

Introduction Petri nets SNP systems Extending SNP systems & Petri nets References

SNP systems & Petri nets SNP systems can benefit greatly from PN theory + literature. They share same ‘look & feel’, among other things. e.g. M.G. Kadjinicolaou, M.B.E Abdelrazik, G. Musgrave, “Structured analysis for neural networks using Petri nets”, Proceedings of the 33rd Midwest Symposium on Circuits and Systems, 2: 70-773, 1990. V.P. Metta, K. Krithivasan, D. Garg, “Spiking Neural P systems and Petri nets”, Proceedings of the International Workshop on Machine Intelligence Research (2009) Francis George Cabarle, Henry Adorna

Some notes on Spiking Neural dP systems and Petri nets

Introduction Petri nets SNP systems Extending SNP systems & Petri nets References

Formal Definition Definition (Petri net) Γ = (P, T , F , W , M0 ) where (1) P = {p1 , p2 , ..., pm } is a finite set of places, (2) T = {t1 , t2 , ..., tn } is a finite set of transitions, (3) F ⊆ (P × T ) ∪ (T × P) is a set of arcs, (4) W : F → {1, 2, 3, ...} is a weight function, (5) M0 : P → {1, 2, 3, ...} is the initial marking P ∩ T = ∅. Francis George Cabarle, Henry Adorna

Some notes on Spiking Neural dP systems and Petri nets

Introduction Petri nets SNP systems Extending SNP systems & Petri nets References

An example

Figure: Simple Petri net example, with 3 places, 2 transitions, and initial marking of (1,0,0)

Francis George Cabarle, Henry Adorna

Some notes on Spiking Neural dP systems and Petri nets

Introduction Petri nets SNP systems Extending SNP systems & Petri nets References

Distributed Petri nets: Shared place A place ps is a shared place between at least two Petri nets Γi and Γj if there exist ti,k ∈ Ti , tj,l ∈ Tj such that there exist (ti,k , ps ), (ps , tj,l ) and there is at least one output arc for ti,k and at least one input arc for tj,l in Γj .

Francis George Cabarle, Henry Adorna

Some notes on Spiking Neural dP systems and Petri nets

Introduction Petri nets SNP systems Extending SNP systems & Petri nets References

Distributed Petri nets: Shared transition A transition ts is a shared transition between at least two Petri nets Γi and Γj if there exist pi,k , pi,k+ω ∈ Pi , pj,l , pj,l+α ∈ Pj , such that there exist (pi,k , ts ), (pj,l , ts ) (input arcs) and (ts , pi,k+ω ), (ts , pj,l+α ) (output arcs), where there is at least one output arc for pi,k , pj,l , and at least one input arcs for pi,k+ω , pj,l+α for some constants α, ω.

Francis George Cabarle, Henry Adorna

Some notes on Spiking Neural dP systems and Petri nets

Introduction Petri nets SNP systems Extending SNP systems & Petri nets References

Distributed Petri net

A distributed Petri net Γ is a Petri net consisting of two or more Petri nets connected via shared places, shared transitions, or both.

Francis George Cabarle, Henry Adorna

Some notes on Spiking Neural dP systems and Petri nets

Introduction Petri nets SNP systems Extending SNP systems & Petri nets References

Distributed Petri nets

Figure: (a) Asynchronous Petri net Γ1 w/ shared place ps . (b) Synchronous Petri net Γ2 w/ shared transition ts based on [Piedrafita et al. 2010]. Francis George Cabarle, Henry Adorna

Some notes on Spiking Neural dP systems and Petri nets

Introduction Petri nets SNP systems Extending SNP systems & Petri nets References

Formal Definition Π = (O, σ1 , . . . , σm , syn, in, out), 1. O = {a}, alphabet of only one object, the system spike a. 2. σ1 , . . . , σm are m neurons of the form: σi = (ni , Ri ), 1 ≤ i ≤ m,: a) ni ≥ 0, initial spike a in neuron σi b) Ri , finite set of rules of w/ 2 forms: (b-1) Spiking rules, E/ac → ap ; d, E is a regular expression over a, p ≥ 1 c ≥ 1, d ≥ 0. (b-2) Forgetting rules, as → λ, s ≥ 1, such that for each rule E/ac → a of type (b-1) from Ri , as ∈ / L(E).

3. syn = {(i, j) | 1 ≤ i, j ≤ m, i 6= j } are synapses between neurons. 4. in, out ∈ {1, 2, . . . , m}, input & output neurons. Francis George Cabarle, Henry Adorna

Some notes on Spiking Neural dP systems and Petri nets

Introduction Petri nets SNP systems Extending SNP systems & Petri nets References

An example

Figure: SNP system Π1 w/ labeled parts. Francis George Cabarle, Henry Adorna

from [Zeng et al. 2011]

Some notes on Spiking Neural dP systems and Petri nets

Introduction Petri nets SNP systems Extending SNP systems & Petri nets References

Distributed SNP systems: SNdP systems ∆ An SNdP system is a construct of the form ∆ = (O, Π1 , Π2 , ..., Πn , esyn) where (1) O = {a} represents the spike, (2) Πi = (O, σi,1 , . . . , σi,mi , syn, ini ) is an SNP system with request rules present only in neuron σini , and each σi,j = (ni,j , Ri,j ) where ni,j is the number of spikes in σi,j and Ri,j is a finite set of rules in σi,j , 1 ≤ j ≤ mi , (3) esyn is a set of external synapses between σi ∈ Πp and σj ∈ Πq and that there is at most one synapse from Πp to Πq and vice versa. Request rule form: E/λ ← ar where E is a regular expression over a and r ≥ 1. [Ionescu et al. 2011] Francis George Cabarle, Henry Adorna

Some notes on Spiking Neural dP systems and Petri nets

Introduction Petri nets SNP systems Extending SNP systems & Petri nets References

SNdP system example ∆1

Francis George Cabarle, Henry Adorna

[Ionescu et al. 2011]

Some notes on Spiking Neural dP systems and Petri nets

Introduction Petri nets SNP systems Extending SNP systems & Petri nets References

Spiking Petri net Ω = (P, T , F , W , D, G, pE , M0 ) where (1) P = {p1 , p2 , ..., pm } is a finite set of places, (2) T = {t1 , t2 , ..., tn } is a finite set of transitions, (3) F ⊆ (P × T ) ∪ (T × P) is a set of arcs, (4) W : F → {1, 2, 3, ...} is a weight function such that W (fi ) denotes a weight of an arc fi ∈ F , (5) D : T → {1, 2, 3, ...} is a delay function such that D(ti ), ti ∈ T , gives a time delay before ti is enabled , (6) G is a guard function G(ti ), ti ∈ T , mapping a boolean type expression to ti which must be true before enabling ti , Francis George Cabarle, Henry Adorna

Some notes on Spiking Neural dP systems and Petri nets

Introduction Petri nets SNP systems Extending SNP systems & Petri nets References

Spiking Petri net

Ω = (P, T , F , W , D, G, pE , M0 ) where (7) pE ∈ P is the place corresponding to the environment of SNP(SNdP) systems and is not part of any sub-Petri net of a distributed Petri net, (8) M0 : P → {1, 2, 3, ...} is the initial marking, such that P ∩ T = ∅. Def’n modified from [Metta et al. 2009b]

Francis George Cabarle, Henry Adorna

Some notes on Spiking Neural dP systems and Petri nets

Introduction Petri nets SNP systems Extending SNP systems & Petri nets References

Modifications + extensions

In [Metta et al. 2009b] [Metta et al. 2009a] pE only has sink transitions (no output arcs). Here we ‘feed’ pE back to the start place of PN. In [Metta et al. 2009b] [Metta et al. 2009a] Spiking PNs (Ω) w/ only 1 input arcs/transition are considered. Here we consider transitions with more than 1 input arc. We convert Ω to SNdP systems (∆) and vice versa.

Francis George Cabarle, Henry Adorna

Some notes on Spiking Neural dP systems and Petri nets

Introduction Petri nets SNP systems Extending SNP systems & Petri nets References

SN ∗ dP system ∆∗

∆∗ = (O, Π1 , Π2 , ..., Πn , eσ, esyn) SNdP system ∆ w/ eσ, set of external neurons (single neuron SNP systems) which cannot have request rules and cannot output to the environment; act only as spike repositories between two or more modules. S A ∆ is a ∆∗ with eσ = ∅, and σi,out = ∅. ∆∗ can have more than one external synapse between Πp , Πq ∈ ∆ ∗

Francis George Cabarle, Henry Adorna

Some notes on Spiking Neural dP systems and Petri nets

Introduction Petri nets SNP systems Extending SNP systems & Petri nets References

Boundedness

PN is k-bounded if each place can have at most k tokens only. An open problem in [Ionescu et al. 2011] (SNdP system introduction) was posted: What if spikes in SNP system environment are limited i.e. no more spikes are produced unless the system produces spikes?

Francis George Cabarle, Henry Adorna

Some notes on Spiking Neural dP systems and Petri nets

Introduction Petri nets SNP systems Extending SNP systems & Petri nets References

Spike bounded SN ∗ dP system: γ-∆∗

A spike bounded SN ∗ dP system γ-∆∗ is an SN ∗ dP system with initially γ finitely many number of spikes in the environment, and further spikes are added only if the system outputs spikes via σout ’s.

Francis George Cabarle, Henry Adorna

Some notes on Spiking Neural dP systems and Petri nets

Introduction Petri nets SNP systems Extending SNP systems & Petri nets References

Theorem 1

Spike bounded SN ∗ dP system to spiking Petri net: For every spike bounded SN ∗ dP system γ-∆∗ there is a corresponding distributed spiking Petri net Ω.

Francis George Cabarle, Henry Adorna

Some notes on Spiking Neural dP systems and Petri nets

Introduction Petri nets SNP systems Extending SNP systems & Petri nets References

Theorem 1 proof idea Each module Πi ∈ γ-∆∗ becomes a sub-Petri net Pi ∈ Ω. Add pE to Ω, M(pE ) = γ. For each σ(i,j) ∈ Πi , add p(i,j) ∈ Pi . For each Spiking rule:

Francis George Cabarle, Henry Adorna

Some notes on Spiking Neural dP systems and Petri nets

Introduction Petri nets SNP systems Extending SNP systems & Petri nets References

Theorem 1 proof idea D(ti ) = 1 G(ti ): if M(pk ) = 2 TRUE, else, FALSE.

Francis George Cabarle, Henry Adorna

Some notes on Spiking Neural dP systems and Petri nets

Introduction Petri nets SNP systems Extending SNP systems & Petri nets References

Theorem 1 proof idea For each Forgetting rule: D(ti ) = 0, G(ti ): if M(pk ) = 2 TRUE, else, FALSE.

Francis George Cabarle, Henry Adorna

Some notes on Spiking Neural dP systems and Petri nets

Introduction Petri nets SNP systems Extending SNP systems & Petri nets References

Theorem 1 proof idea For each Request rule: D(ti ) = 0, G(ti ): if M(pO ) = 2 TRUE, else, FALSE.

Francis George Cabarle, Henry Adorna

Some notes on Spiking Neural dP systems and Petri nets

Introduction Petri nets SNP systems Extending SNP systems & Petri nets References

Thm 1 illustration: recall ∆1 (w.l.og. ∆1 ⇒ γ-∆∗1 , γ = 8)

Francis George Cabarle, Henry Adorna

Some notes on Spiking Neural dP systems and Petri nets

Introduction Petri nets SNP systems Extending SNP systems & Petri nets References

Theorem 1 illustration: γ-∆1 ⇒ Ω1

Francis George Cabarle, Henry Adorna

Some notes on Spiking Neural dP systems and Petri nets

Introduction Petri nets SNP systems Extending SNP systems & Petri nets References

Theorem 2

Spike bounded SN ∗ dP system to spiking Petri net: For every distributed spiking Petri net Ω there is a corresponding spike bounded SN ∗ dP system γ-∆∗ .

Francis George Cabarle, Henry Adorna

Some notes on Spiking Neural dP systems and Petri nets

Introduction Petri nets SNP systems Extending SNP systems & Petri nets References

Theorem 2 proof idea Each module Πi ∈ γ-∆∗ becomes a sub-Spiking PN in Ω. If pE 6= ∅, pE becomes the S environment of γ-∆∗ , γ = M(pE ). If pE = ∅ then σi,out = ∅ and we still create the environment of γ-∆∗ . M(p(i,j) ) = n(i,j) For each ti :

d = D(ti ) Francis George Cabarle, Henry Adorna

Some notes on Spiking Neural dP systems and Petri nets

Introduction Petri nets SNP systems Extending SNP systems & Petri nets References

Theorem 2 proof idea For each ti :

Francis George Cabarle, Henry Adorna

Some notes on Spiking Neural dP systems and Petri nets

Introduction Petri nets SNP systems Extending SNP systems & Petri nets References

Thm 2 illustration: recall Γ1 , Γ2 ⇒ Ω1 , Ω2

Francis George Cabarle, Henry Adorna

Some notes on Spiking Neural dP systems and Petri nets

Introduction Petri nets SNP systems Extending SNP systems & Petri nets References

Theorem 2 illustration: Ω ⇒ γ-∆∗

Francis George Cabarle, Henry Adorna

Some notes on Spiking Neural dP systems and Petri nets

Introduction Petri nets SNP systems Extending SNP systems & Petri nets References

Conclusion Extension of PN (Spiking PNs) and SNP/SNdP systems (SN ∗ dP systems) + conversion from one to another. Gave an idea to an open problem in [Ionescu et al. 2011]: bounded SNdP(SN ∗ dP) systems. Shared place and transition in PNs for synchronous and asynchronous SNP(SN ∗ dP) systems, respectively. What else can we say about power of SN ∗ dP systems? SNP(SN ∗ dP systems) can benefit greatly from theory of PNs: reachability, liveness, etc.

Francis George Cabarle, Henry Adorna

Some notes on Spiking Neural dP systems and Petri nets

Introduction Petri nets SNP systems Extending SNP systems & Petri nets References

References I

˘ [Ionescu et al. 2006] M. Ionescu, Gh. Paun, T. Yokomori, “Spiking Neural P Systems”, Journal Fundamenta Informaticae , vol. 71, issue 2,3 279-308, Feb. 2006. ˘ ´ ´ [Ionescu et al. 2011] M. Ionescu, Gh. Paun, M. Perez-Jim enez, T. Yokomori, “Spiking Neural dP Systems”, Proceedings of the 9th Brainstorming Week on Membrane Computing, RGNC, Sevilla, Spain, pp. 193-208, 2011. [Metta et al. 2009a] V.P. Metta, K. Krithivasan, D. Garg, “Modeling spiking neural P systems using timed Petri nets”, World Congress on Nature and Biologically Inspired Computing pp. 25-30, 2009. dx.doi.org/10.1109/NABIC.2009.5393595

Francis George Cabarle, Henry Adorna

Some notes on Spiking Neural dP systems and Petri nets

Introduction Petri nets SNP systems Extending SNP systems & Petri nets References

References II [Metta et al. 2009b] V.P. Metta, K. Krithivasan, D. Garg, “Spiking Neural P systems and Petri nets”, Proceedings of the International Workshop on Machine Intelligence Research (2009) [Online]. Available http://www.mirlabs.org/nagpur/paper02.pdf [Murata, 1989] T. Murata,“Petri Nets: Properties, analysis and application”, Proceedings of the IEEE, 77(4): pp. 541-580,1989. ˘ ˘ [Paun, 1998] Gh. Paun. “Computing with membranes”. Journal of Computer and System Sciences, 61, 108-143, 1998. ˘ ˘ ´ ´ [Paun et al. 2010] Gh. Paun, M. Perez-Jim enez, “Solving Problems in a Distributed Way in Membrane Computing: dP Systems”, International Journal of Computers Communications & Control, ISSN 1841-9836, 5(2):238-250, 2010.

Francis George Cabarle, Henry Adorna

Some notes on Spiking Neural dP systems and Petri nets

Introduction Petri nets SNP systems Extending SNP systems & Petri nets References

References III

[Piedrafita et al. 2010] R. Piedrafita, D. Tardioli, J.L. Villarroel, “Distributed Implementation of Petri nets in Control Applications”, Petri Nets Applications, Pawel Pawlewski (Ed.) 2010. [Zeng et al. 2011] X. Zeng, H. Adorna, M. A. Mart´ınez-del-Amor, L. Pan, M. ´ ´ Perez-Jim enez, “Matrix Representation of Spiking Neural P Systems”, 11th International Conference on Membrane Computing, Jena, Germany, Aug. 2010 and Lecture Notes in Computer Science, Springer, vol. 6501, 377-39, 2011.

Francis George Cabarle, Henry Adorna

Some notes on Spiking Neural dP systems and Petri nets

Introduction Petri nets SNP systems Extending SNP systems & Petri nets References

Thanks for your attention!

Questions? Francis Cabarle {[email protected]} Henry Adorna {[email protected]} Supported by the DOST-ERDT program Research site: aclab.dcs.upd.edu.ph

Francis George Cabarle, Henry Adorna

Some notes on Spiking Neural dP systems and Petri nets

Some notes on Spiking Neural dP systems and Petri nets

Sep 12, 2011 - A place ps is a shared place between at least two Petri nets Γi and Γj if there exist ti,k ∈ Ti,tj,l ∈ Tj such that there exist. (ti,k ,ps),(ps,tj,l) and ...

1MB Sizes 1 Downloads 199 Views

Recommend Documents

My Notes on Neural Nets for Dual Subspace Pattern Recognition ...
My Notes on Neural Nets for Dual Subspace Pattern Recognition Method.pdf. My Notes on Neural Nets for Dual Subspace Pattern Recognition Method.pdf.

Lending Petri Nets
May 23, 2015 - Introduction. Service-oriented computing (SOC) and cloud computing technologies foster the ... 09124 Cagliari (Italy), e-mail: [email protected].

A Regenerating Spiking Neural Network
allow the design of hardware and software devices capable of re-growing damaged ..... It is interesting to analyse how often a single mutilation can affect the ...

reusing design experience for petri nets through patterns
abstraction than the basic elements. In the software ... terms of basic elements (places, transitions, and arcs), but in terms of more ... of presentation used is not intended to show how to use ..... on computer graphics and artificial intelligence.

Some rough notes on GravitoElectroMagnetism.
I found the GEM equations interesting, and explored the surface of them slightly. Here are some notes, mostly as a reference for myself ... looking at the GEM equations mostly generates questions, especially since I don't have the GR background to un

Debits and credits in Petri nets and linear logic - I.R.I.S. Unica
2 Dipartimento di Informatica, Universit`a di Pisa, Italy. 3 Dipartimento di Matematica, Universit`a degli Studi di Trento, Italy. Abstract. Exchanging resources often ...

pdf-175\artificial-neural-nets-and-genetic-algorithms-proceedings-of ...
... apps below to open or edit this item. pdf-175\artificial-neural-nets-and-genetic-algorithms-p ... nal-conference-in-ales-france-1995-by-david-w-pears.pdf.

Some notes on Besicovitch and Weyl distances over ...
We write H ≤ G if H is a subgroup of G. The classes of the equivalence relation on G defined by xρy iff xy−1 ∈ H are called the right cosets of H. If U is a set of ...

Coloured Petri Nets and CPN Tools for modelling and ... - Springer Link
Mar 13, 2007 - trial-strength computer tool for constructing and anal- ysing CPN models. ... characteristics. Examples of these are business process and workffow modelling [39], manufacturing systems. [11], and agent systems [31]. Examples of industr

Debits and credits in Petri nets and linear logic - I.R.I.S. Unica
After a correct interaction between Alice and Bob, we expect that all re- sources have been consumed: in the logical model, this corresponds to deducing. ΓAlice,ΓBob ⊣ 1 (where ⊣ is the entailment relation of the logic). However, this sequent i

Evolution of spiking neural circuits in ... - Wiley Online Library
Swiss Federal Institute of Technology Lausanne (EPFL), Switzerland. We describe evolution of spiking neural architectures to control navigation of autonomous ...

Saliency extraction with a distributed spiking neural network
Abstract. We present a distributed spiking neuron network (SNN) for ... The bio-inspired paradigm aims at adapting for computer systems what we un- derstand ...

An asynchronous spiking neural network which can ...
I. PROBLEM SPECIFICATION. Our aim is to implement a memory in spiking neurons that can learn any given number of sequences online (a sequence machine) in a single pass or single presentation of the sequence, and predict any learnt sequence correctly.

Covert Attention with a Spiking Neural Network
Neural field. ▻ Spatio-temporal ... Neural field. ▻ Spatio-temporal ... Input Map. Focus Map. Error measure. ▻ Stimulus occupied a spatial position for a time ...

Evolution of Spiking Neural Controllers for Autonomous ...
ing neural controllers for a vision-based mobile robot. All the evolution- ... follow lines [10]. In another line of work, Lewis et al. have developed an analog.

Evolution of spiking neural circuits in autonomous ...
converter is used to reinitialize every 50 ms the pseudorandom number generator ..... Conf on Artificial Intelligence, Denver, Colorado; 1988. pp 118–130. 19.

Covert Attention with a Spiking Neural Network
tions leading to place this work in a bio-inspired framework are explained in 1.1. The properties of the ... approaches offer a good framework for designing efficient and robust methods for extracting .... 5 have been obtained with an desktop Intel.

Some notes on monetary economics Contents
Nov 19, 2015 - 1.6 Affine transfers (“paying interest on large money holdings”) . ... Technology: the consumption good is produced using labor with the .... decreases the real return of this asset, thus making the prospect of working-and-saving-f

some notes on work quality & retirement
Shea (1995) focuses on long-term union contract workers in the PSID for whom the contracts ... equity lines of credit to smooth consumption.3 Second, the size of the disposable income change is relatively large. Although this is ..... sold, credit ca

SOME NOTES ON RANKED STRUCTURES 1 ...
SOME NOTES ON RANKED STRUCTURES. LIANG YU. 1. Inductive definitions and ∆1. 1-boundedness. Let Γ be a map from 2ω to 2ω. Γ is monotonic if X ⊆ Y implies Γ(X) ⊆ Γ(Y ). Γ is progress if X ⊆ Γ(X) for all X. If Γ is progress, we define Î