脉冲神经P系统并行计算的矩阵表示及GPU实现
发布时间:2018-06-25 04:15
本文选题:膜计算 + 脉冲神经P系统 ; 参考:《西华大学》2013年硕士论文
【摘要】:膜计算(又称P系统)是从生命细胞的结构与功能以及组织和器官中细胞群的协作中抽象出来的计算模型。P系统是一类分布式、并行性计算模型。从结构上看,P系统有三种形式:细胞型P系统、组织型P系统和神经型P系统。并行计算特性是P系统的优势之一,对众多应用问题的求解颇具吸引力的。然而,由于当前计算机的串行结构原因,P系统的并行计算还无法真正地模拟或仿真。 GPU(Graphic Processing Unit,图形处理器)是一个相对于CPU的概念,最初的设计理念是为了协助CPU处理图像,它拥有并行处理硬件架构和强大的浮点运算能力,以实现图像处理的硬件加速。如何模拟或仿真各类P系统的并行计算能力是当前膜计算研究的一个热点,因此GPU的出现,特别是其支持矩阵运算的并行实现,,为该研究提供了一个新的途径。 本文主要选取两种脉冲神经P系统,实现其矩阵表示并给出了它们的GPU实现算法。详细的研究工作如下: (1)研究并提出了耗尽型脉冲神经P系统并行计算的矩阵表示。根据这个矩阵表示,给出了耗尽型脉冲神经P系统的GPU实现算法。几个示例的仿真结果说明了其GPU实现的可行性。 (2)针对时延脉冲神经P系统,提出了其并行计算的矩阵表示,并进一步研究了了GPU实现算法。通过几个示例的仿真,验证了延时脉冲神经P系统并行计算的GPU实现的可行性和有效性。
[Abstract]:Membrane computing (also called P system) is a computing model abstracted from the structure and function of living cells and the cooperation of cell groups in tissues and organs. P system is a kind of distributed parallel computing model. There are three types of P system: cellular P system, tissue P system and nerve P system. Parallel computing is one of the advantages of P system, and it is attractive to solve many application problems. However, due to the current serial structure of computers, parallel computing in P system can not be really simulated or simulated. GPU (graphic processing Unit) is a concept relative to CPU. The original design idea is to assist CPU in image processing. It has parallel processing hardware architecture and powerful floating-point computing ability to achieve hardware acceleration of image processing. How to simulate or simulate the parallel computing capability of various P systems is a hot topic in the research of membrane computing. Therefore, the appearance of GPU, especially the parallel implementation of matrix operations, provides a new way for the research. In this paper, two kinds of impulsive neural P systems are selected to realize their matrix representation and their GPU implementation algorithms are given. The detailed research work is as follows: (1) the matrix representation of parallel computation for depleted impulsive neural P systems is studied and proposed. According to this matrix representation, a GPU implementation algorithm for depleted impulsive neural P system is presented. The simulation results of several examples show the feasibility of the GPU implementation. (2) for the delay impulsive neural P system, the matrix representation of parallel computation is proposed, and the GPU implementation algorithm is further studied. Through several examples, the feasibility and effectiveness of parallel GPU implementation for delayed impulsive neural P system are verified.
【学位授予单位】:西华大学
【学位级别】:硕士
【学位授予年份】:2013
【分类号】:TP38;TP391.41
【参考文献】
相关期刊论文 前6条
1 张葛祥;潘林强;;自然计算的新分支——膜计算[J];计算机学报;2010年02期
2 兰丽;;一种新型膜计算方法在并行排序中的应用[J];兰州交通大学学报;2011年04期
3 黄春毅;董笑菊;龙环;;用P系统解决排序问题[J];上海交通大学学报;2008年02期
4 詹云;赵新灿;谭同德;;基于OpenCL的异构系统并行编程[J];计算机工程与设计;2012年11期
5 邵杰;彭宏;蒋洋;杨雨凡;黄小丽;张加容;高雪峰;;耗尽型脉冲神经P系统的矩阵表示及GPU实现[J];西华大学学报(自然科学版);2013年02期
6 韩俊刚;刘有耀;张晓;;图形处理器的历史现状和发展趋势[J];西安邮电学院学报;2011年03期
本文编号:2064523
本文链接:https://www.wllwen.com/kejilunwen/jisuanjikexuelunwen/2064523.html