computationally complete spiking neural p systems without
play

Computationally Complete Spiking Neural P Systems Without Delay: - PowerPoint PPT Presentation

Computationally Complete Spiking Neural P Systems Without Delay: Two Types of Neurons Are Enough Rudolf Freund 1 , Marian Kogler 1 , 2 1 Faculty of Informatics, Vienna University of Technology, Austria 2 Institute of Computer Science, Martin


  1. Computationally Complete Spiking Neural P Systems Without Delay: Two Types of Neurons Are Enough Rudolf Freund 1 , Marian Kogler 1 , 2 1 Faculty of Informatics, Vienna University of Technology, Austria 2 Institute of Computer Science, Martin Luther University Halle-Wittenberg, Germany Email: rudi@emcc.at , marian@emcc.at , kogler@informatik.uni-halle.de CMC11

  2. Overview Preliminaries Results Suggestions for Future Work

  3. Register Machines Definition (Register Machine) A register machine is a construct M = ( n , B , p 0 , p h , I ) where 1. n , n ≥ 1, is the number of registers, 2. B is the set of instruction labels, 3. p 0 is the start label, 4. p h is the halting label (only used for the HALT instruction) and 5. I is a set of (labeled) instructions.

  4. Register Machines Definition (Register Machine (ctd.)) Instructions are of the following forms: ◮ p i : ( ADD ( r ) , p j , p k ) increments the value in register r and continues with one of the instructions labeled by p j and p k , chosen in a nondeterministic way, ◮ p i : ( SUB ( r ) , p j , p k ) tries to decrement the value in register r ; if the register was non-empty before the instruction, the computation continues with the instruction labeled with p j , if not, it continues with the instruction p k ; ◮ p h : HALT halts the machine.

  5. Register Machines Definition (Register Machine (ctd.)) ◮ Deterministic register machines can be constructed by imposing the condition p j = p k on ADD -instructions. ◮ We will be using nondeterministic register machines as generators and deterministic register machines as acceptors. ◮ Every recursively enumerable set of (vectors of) natural numbers with k components can be generated with only k + 2 registers, where the first k registers are never decremented (Minsky, 1967).

  6. Spiking Neural P Systems (Without Delays) Definition (Spiking neural P system) A spiking neural P system (without delays) is a construct Π = ( O , ρ 1 , ..., ρ n , syn , in , out ) where 1. O = { a } is the (unary) set of objects (the object a is called spike ), 2. ρ 1 , ..., ρ n are the neurons, where ρ i = ( d i , R i ) for 1 ≤ i ≤ n , with d i being the initial configuration of the neuron i and R i being the set of rules, 3. in is the input neuron (with the only function to spike once in generating spiking neural P systems in order to start a computation), and 4. out is the output neuron (no function in accepting spiking neural P systems).

  7. Spiking Neural P Systems (Without Delays) Definition (Spiking neural P system (ctd.)) Possible forms for the rules: ◮ E / a i → a j , where E is a regular expression over O and i , j ≥ 1 ( firing rules ) or ◮ a i → λ , where i ≥ 1 ( forgetting rules ). There must not be any rule a i → λ such that a i ∈ L ( E ) for some E of a firing rule. ◮ syn ⊆ { 1 , ..., n } × { 1 , ..., n } are the synapses, where ( i , j ) ∈ syn indicates a synapse from i to j ,

  8. Spiking Neural P Systems (Without Delays) Definition (Spiking neural P system (ctd.)) ◮ A spiking neural P system inputs and outputs numbers via a spike train. A spike train starts with a spike given in step t 1 and ends with a spike given in step t 2 . The number is specified by t 2 − t 1 − 1, i.e., the number of steps that elapse between the two spikes. It accepts an input by a series of configurations, starting from the initial configuration and ending in a halting configuration. ◮ Rules of the form E / a i → a j where L ( E ) is finite (infinite) are called bounded (unbounded) rules . ◮ Two neurons ρ i and ρ j are of the same type if and only if R i = R j , d i = d j and |{ ( i , k ) ∈ sym | k ∈ { 1 , ..., n }}| = |{ ( j , k ) ∈ sym | k ∈ { 1 , ..., n }}| .

  9. Results ◮ Accepting spiking neural P systems (without delays) with only two types of neurons are computationally complete. ◮ Generating spiking neural P systems without delays with only two types of neurons are computationally complete. ◮ Corollary: Spiking neural P systems without delays with only three neurons with unbounded rules are computationally complete.

  10. Results ◮ Accepting spiking neural P systems (without delays) with only two types of neurons are computationally complete. ◮ Generating spiking neural P systems without delays with only two types of neurons are computationally complete. ◮ Corollary: Spiking neural P systems without delays with only three neurons with unbounded rules are computationally complete.

  11. The Two Types λ λ a / a → a a → λ a 2 → λ a 3 ( a 2 ) ∗ / a 3 → a a 3 / a 3 → a a 4 / a 4 → a a 5 → λ Type 1 Type 2

  12. A Dummy Structure d 1 � ❅ � ✠ ❅ ❘ d 2 d 3 ❆ ❑ ❆ ❅ � ✁ ✁ ✕ ❆ ❆ ❅ ❘ � ✠ ✁ ✁ ❆ ❆ ✁ ✁ ❆ ❆ ✁ ✁ ❆ ❆ ✁ ✁ ❆ ❆ d 4 ✁ ✁ ❆ ❆ ✁ ✁ ❆ ❆ ❯ ☛ ✁ ✁ d 5

  13. A Dummy Structure a d 1 � ❅ � ✠ ❅ ❘ d 2 d 3 ❆ ❑ ❆ ❅ � ✕ ✁ ✁ ❆ ❆ ❅ ❘ ✠ � ✁ ✁ ❆ ❆ ✁ ✁ ❆ ❆ ✁ ✁ ❆ ❆ ✁ ✁ ❆ ❆ d 4 ✁ ✁ ❆ ❆ ✁ ✁ ❆ ❆ ❯ ✁ ☛ ✁ d 5

  14. A Dummy Structure d 1 � ❅ � ✠ ❘ ❅ a a d 2 d 3 ❆ ❆ ❑ ❅ � ✁ ✕ ✁ ❆ ❆ ❘ ❅ ✠ � ✁ ✁ ❆ ❆ ✁ ✁ ❆ ❆ ✁ ✁ ❆ ❆ ✁ ✁ ❆ ❆ d 4 ✁ ✁ ❆ ❆ ✁ ✁ ❆ ❯ ❆ ✁ ☛ ✁ d 5

  15. A Dummy Structure d 1 � ❅ � ✠ ❅ ❘ d 2 d 3 ❆ ❑ ❆ ❅ � ✁ ✁ ✕ ❆ ❆ ❘ ❅ ✠ � ✁ ✁ ❆ ❆ ✁ ✁ ❆ ❆ ✁ ✁ aa ❆ ❆ ✁ ✁ ❆ ❆ d 4 ✁ ✁ ❆ ❆ ✁ ✁ ❆ ❯ ❆ ✁ ☛ ✁ aa d 5

  16. A Dummy Structure d 1 � ❅ � ✠ ❅ ❘ d 2 d 3 ❆ ❑ ❆ ❅ � ✁ ✁ ✕ ❆ ❆ ❅ ❘ � ✠ ✁ ✁ ❆ ❆ ✁ ✁ ❆ ❆ ✁ ✁ ❆ ❆ ✁ ✁ ❆ ❆ d 4 ✁ ✁ ❆ ❆ ✁ ✁ ❆ ❆ ❯ ☛ ✁ ✁ d 5

  17. Simulating an ADD -instruction p i 1 � ❅ � ✠ ❅ ❘ p i 2 p i 3 ❅ � ❅ ❘ � ✠ r ✛✘ ✛✘ ❄ ❄ p j D ✚✙ ✚✙

  18. Simulating an ADD -instruction a p i 1 � ❅ � ✠ ❅ ❘ p i 2 p i 3 ❅ � ❘ ❅ � ✠ r ✛✘ ✛✘ ❄ ❄ p j D ✚✙ ✚✙

  19. Simulating an ADD -instruction p i 1 � ❅ � ✠ ❘ ❅ a a p i 2 p i 3 ❅ � ❘ ❅ � ✠ r ✛✘ ✛✘ ❄ ❄ p j D ✚✙ ✚✙

  20. Simulating an ADD -instruction p i 1 � ❅ � ✠ ❅ ❘ p i 2 p i 3 ❅ � ❅ ❘ � ✠ aa r ✛✘ ✛✘ ❄ ❄ p j D ✚✙ ✚✙ a a

  21. Simulating a SUB -instruction ... ... p 1 1 p i 1 p n 1 ❄ ❄ ... ... p i 2 p i 3 ❍❍❍❍ ❄ ❄ ❥ ... ... p i 4 p i 5 r ❄ ❄ ❄ ... ... ... ... ... ˜ p i 1 p i 2 ˜ p ′ p ′ p ′ � 1 1 i 1 n 1 � � ✠ ❄ ❄ ❄ ❄

  22. Simulating a SUB -instruction ... � � ✠ � ❄ ❄ ❄ ❄ ... ... p ′ p ′ p i 3 ˜ ˜ p i 4 ˜ p i 5 i 2 i 3 ❅ � ❅ � ❅ � ❅ ❘ ❄❄ � ✠ p ′ i s ✛✘ ❄ p i s ✚✙ ✲ ✛ p ′ i f ✛✘ ❄ p i f ✚✙

  23. Simulating a SUB -instruction ... ... a p 1 1 p i 1 p n 1 ❄ ❄ ... ... p i 2 p i 3 ❍❍❍❍ ❄ ❄ ❥ ... ... p i 4 p i 5 aa r ❄ ❄ ❄ ... ... ... ... ... ˜ p i 1 ˜ p i 2 p ′ p ′ p ′ � 1 1 i 1 n 1 � � ✠ ❄ ❄ ❄ ❄

  24. Simulating a SUB -instruction ... ... p 1 1 p i 1 p n 1 ❄ ❄ ... ... a a p i 2 p i 3 ❍❍❍❍ ❄ ❄ ❥ ... ... p i 4 p i 5 aa r ❄ ❄ ❄ ... ... ... ... ... ˜ p i 1 p i 2 ˜ p ′ p ′ p ′ � 1 1 i 1 n 1 � � ✠ ❄ ❄ ❄ ❄

  25. Simulating a SUB -instruction ... ... p 1 1 p i 1 p n 1 ❄ ❄ ... ... p i 2 p i 3 ❍❍❍❍ ❄ ❄ ❥ ... ... a a p i 4 p i 5 aaa r ❄ ❄ ❄ ... ... ... ... ... ˜ p i 1 p i 2 ˜ p ′ p ′ p ′ � 1 1 i 1 n 1 � � ✠ ❄ ❄ ❄ ❄

  26. Simulating a SUB -instruction ... ... p 1 1 p i 1 p n 1 ❄ ❄ ... ... p i 2 p i 3 ❍❍❍❍ ❄ ❄ ❥ ... ... p i 4 p i 5 r ❄ ❄ ❄ ... ... ... ... ... a a a a a ˜ p i 1 p i 2 ˜ p ′ p ′ p ′ � 1 1 i 1 n 1 � � ✠ ❄ ❄ ❄ ❄

  27. Simulating a SUB -instruction ... � � ✠ � ❄ ❄ ❄ ❄ ... ... a a a a a p ′ p ′ p i 3 ˜ p i 4 ˜ p i 5 ˜ i 2 i 3 ❅ � ❅ � ❅ � ❅ ❘ ❄❄ � ✠ p ′ i s ✛✘ ❄ p i s ✚✙ ✲ ✛ p ′ i f ✛✘ ❄ p i f ✚✙

  28. Simulating a SUB -instruction ... � � ✠ � ❄ ❄ ❄ ❄ ... ... p ′ p ′ p i 3 ˜ p i 4 ˜ ˜ p i 5 i 2 i 3 ❅ � ❅ � ❅ � ❅ ❘ ❄❄ � ✠ aa p ′ aa i s ✛✘ ❄ p i s ✚✙ aa ✲ ✛ aa p ′ a i f ✛✘ ❄ p i f ✚✙

  29. Simulating a SUB -instruction ... � � ✠ � ❄ ❄ ❄ ❄ ... ... p ′ p ′ p i 3 ˜ ˜ p i 4 p i 5 ˜ i 2 i 3 ❅ � ❅ � ❅ � ❅ ❘ ❄❄ � ✠ p ′ i s ✛✘ ❄ p i s ✚✙ a ✲ ✛ p ′ i f ✛✘ ❄ p i f ✚✙

  30. Simulating a SUB -instruction ... ... a p 1 1 p i 1 p n 1 ❄ ❄ ... ... p i 2 p i 3 ❍❍❍❍ ❄ ❄ ❥ ... ... p i 4 p i 5 r ❄ ❄ ❄ ... ... ... ... ... ˜ p i 1 p i 2 ˜ p ′ p ′ p ′ � 1 1 i 1 n 1 � � ✠ ❄ ❄ ❄ ❄

Recommend


More recommend