• 主题
高级搜索  >
历史搜索 清空历史
首页> 外文会议>2011 19th Iranian Conference on Electrical Engineering >A GPU based simulation of multilayer spiking neural networks
【24h】

A GPU based simulation of multilayer spiking neural networks

机译：基于GPU的多层尖峰神经网络仿真

获取原文
页面导航
  • 摘要
  • 著录项
  • 相似文献
  • 相关主题

摘要

Nowadays, despite significant advances in VLSI technology, in the case of massively parallel systems still new computational architectures are required. Using graphic processing units (GPU) as a low-cost and high performance computing platform is an efficient preferred approach to such problems. Simulation of spiking neural networks (SNN) is a well-known challenge encountering these barriers. In this paper we demonstrate an Izhikevich neuron simulator that runs on a single GPU. The GPU-SNN model (running on an NVIDIA GT325M with 1GB of memory) is up to 11 times faster than a CPU version when more than one million neurons with 75 billion synaptic connections. Simulation results are compared for different single GPU with the CPU based simulation different single GPU. Simulation method is based on a new method of virtual synaptic computation, which performs the calculation with low memory usage.
机译：如今，尽管VLSI技术取得了重大进步，但在大规模并行系统的情况下，仍然需要新的计算架构。使用图形处理单元（GPU）作为低成本和高性能的计算平台是解决此类问题的有效首选方法。尖峰神经网络（SNN）的仿真是遇到这些障碍的众所周知的挑战。在本文中，我们演示了在单个GPU上运行的Izhikevich神经元模拟器。当超过一百万个具有750亿个突触连接的神经元时，GPU-SNN模型（运行在具有1GB内存的NVIDIA GT325M上）比CPU版本快11倍。将不同的单个GPU的仿真结果与基于CPU的不同的单个GPU的仿真结果进行比较。仿真方法基于虚拟突触计算的新方法，该方法以较低的内存使用量执行计算。

著录项

相似文献

  • 外文文献
  • 中文文献
  • 专利
获取原文

客服邮箱：kefu@zhangqiaokeyan.com

京公网安备：11010802029741号 ICP备案号：京ICP备15016152号-6 六维联合信息科技 (北京) 有限公司©版权所有

  • 客服微信

  • 服务号