Multiplexed spatiotemporal communication model in artificial neural networks.
Multiplexed spatiotemporal communication model in artificial neural networks.
复制标题
人工神经网络中的多路时空通信模型。
DOI：
10.11648/j.acis.20130106.11
复制
发表时间：
2014
期刊：
影响因子：
--
通讯作者：
Y-W.
中科院分区：
文献类型：
--
作者：
Tamura;S.;Nishitani;Y.;Kamimura;T.;Hosokawa;C.;Miyoshi;T.;Sawai;H.;Mizuno-Matsumoyo;Y.;CHen;Y-W.
众所周知，大脑的不同区域之间存在着相互联系。然而，到目前为止，通信规则还没有成功地分析。来自神经元细胞的尖峰序列已经被简单地视为具有相应神经元细胞的激活水平的密度调制波，或者，至多，它们已经使用序列之间的传统度量进行了分析。来自神经元细胞的尖峰序列具有类似随机的模式，几乎没有提供关于编码规则的线索。在一个随机产生的由阈值元件组成的人工3× 3多路时空通信神经网络中，我们发现在模拟过程中产生了伪随机序列，类似于由培养的大鼠大脑神经网络产生的随机序列。仿真中的瞬态序列模式被认为是电路结构的反映。这些随机形状的电路产生伪随机序列，用作多路复用通信的代码。虽然目前电路权重是随机生成的，但将有可能扩展这种方法以通过学习来确定网络权重。本文提供的模拟结果支持培养神经网络的研究结果。
It is well known that there is intercommunication among the different areas of the brain. However, till date, the rules of communication have not been successfully analyzed. The spike trains from neuronal cells have been simply treated as density-modulated waves with an activation level of the corresponding neuronal cells, or, at most, they have been analyzed using traditional metrics between sequences. The spike trains from neuronal cells have a random-like pattern that provides few clues regarding a coding rule. Here in a randomly generated artificial 3× 3 multiplexed spatiotemporal communication neural network composed of threshold elements, we showed that pseudorandom sequences were generated during the simulation, similar to the random sequences generated by the cultured neural network of the rat brain. The transiently generated sequence patterns in the simulation were regarded as reflecting the circuit structure. These randomly shaped circuits generated pseudorandom sequences that functioned as codes for multiplexing communication. Although the circuit weights are randomly generated at present, it will be possible to extend this approach to determine the network weights by learning. This paper provides simulation results that support findings on cultured neural network.