ABSTRACT
The parallel simulation of Spiking Neural P systems is mainly based on a matrix representation, where the graph inherent to the neural model is encoded in an adjacency matrix. The simulation algorithm is based on a matrix-vector multiplication, which is an operation efficiently implemented on parallel devices. However, when the graph of a Spiking Neural P system is not fully connected, the adjacency matrix is sparse and hence, lots of computing resources are wasted in both time and memory domains. For this reason, two compression methods for the matrix representation were proposed in a previous work, but they were not implemented nor parallelized on a simulator. In this paper, they are implemented and parallelized on GPUs as part of a new Spiking Neural P system with delays simulator. Extensive experiments are conducted on high-end GPUs (RTX2080 and A100 80GB), and it is concluded that they outperform other solutions based on state-of-the-art GPU libraries when simulating Spiking Neural P systems.
Subject(s)
Action Potentials , Algorithms , Computer Graphics , Models, Neurological , Action Potentials/physiology , Neurons/physiology , Neural Networks, Computer , Computer Simulation , HumansABSTRACT
Spiking neural P systems (SNP systems) are parallel and non-deterministic models of computation, inspired by the neural system of the brain. A variant of SNP systems known as SNP systems with structural plasticity (SNPSP systems) includes the feature of adding or removing synapses among neurons. This feature is inspired by plasticity from neuroscience during cognition and learning. Despite the reductionist framework of SNP and SNPSP systems, such as brain-like systems are capable of computational universality. In particular, we use SNPSP systems in this paper to compute some classes of languages from the Chomsky hierarchy: FIN, REG, and RE. The computations of such classes continue a research direction established in the previous paper. We also emphasize the (dis)advantages of synapse plasticity in the neural system, compared with existing features of SNP systems, when generating languages.
Subject(s)
Action Potentials/physiology , Models, Neurological , Neural Networks, Computer , Neuronal Plasticity/physiology , Programming Languages , Computers, Molecular , Synapses/physiologyABSTRACT
Spiking neural P systems (SN P systems) are models of computation inspired by biological spiking neurons. SN P systems have neurons as spike processors, which are placed on the nodes of a directed and static graph (the edges in the graph are the synapses). In this paper, we introduce a variant called SN P systems with scheduled synapses (SSN P systems). SSN P systems are inspired and motivated by the structural dynamism of biological synapses, while incorporating ideas from nonstatic (i.e., dynamic) graphs and networks. In particular, synapses in SSN P systems are available only at specific durations according to their schedules. The SSN P systems model is a response to the problem of introducing durations to synapses of SN P systems. Since SN P systems are in essence static graphs, it is natural to consider them for dynamic graphs also. We introduce local and global schedule types, also taking inspiration from the above-mentioned sources. We prove that SSN P systems are computationally universal as number generators and acceptors for both schedule types, under a normal form (i.e., a simplifying set of restrictions). The introduction of synapse schedules for either schedule type proves useful in programming the system, despite restrictions in the normal form.