项目名称: 脉冲神经网络敏感性及其应用研究
项目编号: No.61503031
项目类型: 青年科学基金项目
立项/批准年度: 2016
项目学科: 其他
项目作者: 杨静
作者单位: 北京师范大学珠海分校
项目金额: 18万元
中文摘要: 本项目的研究对象是第三代人工神经网络:脉冲神经网络(Spiking neural network)的敏感性及其应用。脉冲神经网络的输入输出为具有时间特性的脉冲序列,其运行机制相比其他人工神经网络更加接近于生物神经网络,也更为复杂。本项目将首先从spiking神经网络的敏感性研究入手,分析各种不同的输入参数扰动对于网络输出的可能影响,并给出相应的量化算法。然后以此为基础,深入研究网络可调参数与其行为表现之间的内在联系。最后,利用所得到的量化敏感性为衡量工具指导spiking网络的结构优化,并对spiking神经网络的脉冲序列学习算法加以改进。.本项目的研究意义就在于它将在很大程度上促进人工神经网络的发展,使得人工神经网络拥有更进一步的智能,为科学进步和社会发展发挥更大的作用。
中文关键词: 脉冲神经网络;脉冲神经元;敏感性;学习算法
英文摘要: The research object of this project is the third generation model of artificial neural networks, namely, the Spiking Neural Networks(SNNs). The input and output of SNNs are spike trains with time character.The running mechanism of SNNs is more similar to the operating mode of the biological nervous system and more complex than other traditional artificial neural networks.This project will firstly study the sensitivity of SNNs. We will analyze the possible influence to network's output caused by different input perturbations and propose the corresponding quantified algorithm. Secondly, we will study the inner relationship between SNNs’ adaptive parameters and their behaviors based on the previous work. Finally, we will use the quantified sensitivity as tool to optimize the network structure and improve the learning mechanism of SNNs..The significance of this research is that it could futher promote development of the research on artifical neural networks, make artifical neural networks possess more human intelligence, and contribute more in scientific progress and social development.
英文关键词: Spiking neural network;Spiking neuron;Sensitivity;Learning algrithm