Identification of neural network structure from multiple spike sequences
Identification of neural network structure from multiple spike sequences
复制标题
从多个尖峰序列中识别神经网络结构
DOI：
10.1007/978-3-642-34481-7_23
复制
发表时间：
2012
期刊：
影响因子：
--
通讯作者：
Tohru Ikeguchi
中科院分区：
文献类型：
--
作者：
Kaori Kuroda;Kantaro Fujiwara;Tohru Ikeguchi
本文提出了一种新的仅由多个脉冲序列估计神经网络中神经元间连通性方向的方法。所提出的方法是基于尖峰时间度量，或统计措施，以量化两个尖峰序列之间的不相似程度，和偏化分析。为了解决这个问题，我们修改了尖峰时间度量中传统成本的定义。然后，该方法可以有效地估计神经元之间的连接方向。为了验证该方法的有效性，我们将该方法应用于由数学神经网络模型产生的多个尖峰序列。因此，我们的方法可以估计神经网络的结构和耦合的方向与高精度。
In this paper, we propose a new estimation method of direction of the connectivity between neurons in neural network only from multiple spike sequences. The proposed method is based on the spike time metric, or a statistical measure to quantify a degree of dissimilarity between two spike sequences, and the partialization analysis. To resolve this issue, we modify the definition of the conventional cost in the spike time metric. Then, the proposed method can effectively estimate direction of connectivity between neurons. To check the validity, we applied the proposed method to multiple spike sequences that are produced by a mathematical neural network model. As a result, our method can estimate the neural network structure and the direction of couplings with high accuracy.