Search

Methods for optimizing and evolving spiking neural P systems (in short, SN P systems) have been previously developed with the use of a genetic algorithm framework. So far, these computations, both evolving and simulating, were done only sequentially. Due to the non-deterministic and parallel nature of SN P systems, it is natural to harness parallel processors in implementing its evolution and simulation. In this work, a parallel framework for the evolution of SN P Systems is presented. This is the result of extending our previous work by implementing it on a CUDA-enabled graphics processing unit and adapting CuSNP design in simulations. Using binary addition and binary subtraction with 3 different categories each as initial SN P systems, the GPU-based evolution runs up to 9x faster with respect to its CPU-based evolution counterparts. Overall, when considering the whole process, the GPU framework is up to 3 times faster than the CPU version.

To date, parallel simulation algorithms for spiking neural P (SNP) systems are based on a matrix representation. This way, the simulation is implemented with linear algebra operations, which can be easily parallelized on high performance computing platforms such as GPUs. Although it has been convenient for the first generation of GPU-based simulators, such as CuSNP, there are some bottlenecks to sort out. For example, the proposed matrix representations of SNP systems lead to very sparse matrices, where the majority of values are zero. It is known that sparse matrices can compromise the performance of algorithms since they involve a waste of memory and time. This problem has been extensively studied in the literature of parallel computing. In this paper, we analyze some of these ideas and apply them to represent some variants of SNP systems. We also provide a new simulation algorithm based on a novel compressed representation for sparse matrices. We also conclude which SNP system variant better suits our new compressed matrix representation.

In this paper, we present a genetic algorithm framework for evolving Spiking Neural P Systems with rules on synapses (RSSNP systems, for short). Starting with an initial RSSNP system, we use the genetic algorithm framework to obtain a derived RSSNP system with fewer resources (fewer and simpler rules, fewer synapses, less initial spikes) that can still produce the expected output spike trains. Different methods in the selection of parents and in the calculation of fitness are incorporated. We also try the framework on 5 RSSNP systems that compute bitwise AND, OR, NOT, ADD, and SUB respectively to gather data on how the framework behaves. Lastly, we discuss the asymptotic complexity of the algorithm and its effectiveness in generating fitter RSSNP systems based on which methods were used.

Spiking Neural P systems (in short, SNP systems) are computing models based on living neurons. SNP systems are non-deterministic and parallel, hence making use of a parallel processor such as a graphics processing unit (in short, GPU) is a natural candidate for simulations. Matrix representations and algorithms were previously developed for simulating SNP systems. In this work, our two results extend previous works in simulating SNP systems in the GPU: (a) the number of neurons the simulator can handle is now arbitrary; (b) SNP systems are now represented in a dense instead of sparse way. The impact in terms of time and space of these extensions to the GPU simulator are analysed. As expected, SNP systems with more neurons need more simulation time, although the simulator performance can scale (i.e. perform better) with larger GPUs. The dense representation helps in the simulation of larger systems.

Current parallel simulation algorithms for Spiking Neural P (SNP) systems are based on a matrix representation. This helps to harness the inherent parallelism in algebraic operations, such as vector-matrix multiplication. Although it has been convenient for the rst parallel simulators running on Graphics Processing Units (GPUs), such as CuSNP, there are some bottlenecks to cope with. For example, matrix representation of SNP systems with a low-connectivity-degree graph lead to sparse matrices, i.e. containing more zeros than actual values. Having to deal with sparse matrices downgrades the performance of the simulators because of wasting memory and time. However, sparse matrices is a known problem on parallel computing with GPUs, and several solutions and algorithms are available in the literature. In this paper, we brie y analyse some of these ideas and apply them to represent some variants of SNP systems. We also conclude which variant better suit a sparse-matrix representation.

Spiking Neural P system (or SN P system) is a computing model based on the neurons in a living being. It is composed of neurons containing spikes interconnected by synapses. Each neuron contain a set of rules which will determine how the spikes are passed in the system. It is a non-deterministic and parallel system which makes GPU a good candidate for simulating this computing model. A matrix representation for system without delay was previously developed and an algorithm for simulating deterministic systems with delay was also presented. In this work, an algorithm for simulating non-deterministic Spiking Neural P System was presented. To accelerate simulations of Spiking Neural P Systems, this algorithm was then implemented and used to simulate nonuniform and uniform solution to the subset sum problem as a case study. Time and space resources in the GPU of such simulations are then compared and analyzed

Spiking neural P systems (in short, SN P systems) are models of computation inspired by biological neurons. CuSNP is a project involving sequential (CPU) and parallel (GPU) simulators for SN P systems. In this work, we report the following results: a P-Lingua le parser is included, for ease of use when performing simulations; extension of the matrix representation of SN P systems to include delay; comparison and analysis of our simulators by simulating two types (bitonic and generalized) of parallel sorting networks; extension of supported types of regular expressions in SN P systems. Our GPU simulator is better suited for generalized sorting as compared to bitonic sorting networks, and the GPU simulators run up to 50 faster than our CPU simulator. Finally, we discuss our experiments and provide directions for further work.

Spiking neural P systems (in short, SN P systems) are parallel models of computations inspired by the spiking ( ring) of biological neurons. In SN P systems, neurons function as spike processors and are placed on nodes of a directed graph. Synapses, the connections between neurons, are represented by arcs or directed endges in the graph. Not only do SN P systems have parallel semantics (i.e. neurons operate in parallel), but their structure as directed graphs allow them to be represented as vectors or matrices. Such representations allow the use of linear algebra operations for simulating the evolution of the system con gurations, i.e. computations. In this work, we continue the implementations of SN P systems with delays, i.e. a delay is associated with the sending of a spike from a neuron to its neighbouring neurons. Our implementation is based on a modi ed representation of SN P systems as vectors and matrices for SN P systems without delays. We us massively parallel processors known as graphics processing units (in short, GPUs) from NVIDIA. For experimental validation, we use SN P systems implementing generalized sorting networks. We report a speedup, i.e. the ratio between the running time of the sequential over the parallel simulator, of up to approximately 51 times for a 512-size input to the sorting network.

Spiking neural P systems (in short, SNP systems) are parallel, distributed, and nondeterministic computing devices inspired by biological spiking neurons. Recently, a class of SNP systems known as SNP systems with structural plasticity (in short, SNPSP systems) were introduced. SNPSP systems represent a class of SNP systems that have dynamism applied to the synapses, i.e. neurons can use plasticity rules to create or remove synapses. In this work, we impose the restriction of sequentiality on SNPSP systems, using four modes: max, min, maxpseudo, and min-pseudo sequentiality. We also impose a normal form for SNPSP systems as number acceptors and generators. Conditions for (non)universality are then provided. Speci cally, acceptors are universal in all modes, while generators need a nondeterminism source in two modes, which in this work is provided by the plasticity rules.

Spiking neural P systems (in short, SN P systems) are membrane computing models inspired by the pulse coding of information in biological neurons. SN P systems with standard rules have neurons that emit at most one spike (the pulse) each step, and have either an input or output neuron connected to the environment. A variant known as SN P modules generalize SN P systems by using extended rules (more than one spike can be emitted each step) and a set of input and output neurons. In this work we continue relating SN P modules and finite automata. In particular, we amend and improve previous constructions for the simulatons of deterministic finite automata and state transducers. Our improvements reduce the number of neurons from three down to one, so our results are optimal. We also simulate finite automata with output, and we use these simulations to generate automatic sequences.

Spiking neural P systems (in short, SN P systems) are models of computation inspired by biological neurons. In this work, we report our ongoing e orts to improve simulators for SN P systems. CuSNP is a project involving sequential and parallel simulators, and in this work we include a PLingua le parser. The PLingua le parser is for ease of use when performing simulations to be executed either in the CPU or in CUDA graphics processing units (in short, GPUs). Our results also include a comparison and analysis of the simulator we developed by simulating two types of parallel soring networks: generalized and bitonic. At present, our GPU simulator is better suited on the former type based on the pro ling of our GPU kernel functions, i.e. our GPU simulators run up to 50 faster than the sequential simulator but simulations of bitonic networks run slightly slower than generalized networks. We also implemented an algorithm based on nite automata to allow more forms of regular expressions in the simulated SN P systems.

Spiking neural P systems with structural plasticity (in short, SNPSP systems) are models of computations inspired by the function and structure of biological neurons. In SNPSP systems, neurons can create or delete synapses using plasticity rules. We report two families of solutions: a non-uniform and a uniform one, to the NP-complete problem Subset Sum using SNPSP systems. Instead of the usual rule-level nondeterminism (choosing which rule to apply) we use synapse-level nondeterminism (choosing which synapses to create or delete). The nondeterminism due to plasticity rules have the following improvements from a previous solution: in our non-uniform solution, plasticity rules allowed for a normal form to be used (i.e. without forgetting rules or rules with delays, system is simple, only synapse-level nondeterminism); in our uniform solution the number of neurons and the computation steps are reduced.