一种简单有效的脉冲串概率模型
首发时间:2020-09-11
摘要:对神经元脉冲发放规律的准确估计是使脉冲神经网络能更有效地模拟生物智能的基础,对脉冲序列的建模工作长期以来通过引入更多控制因素、应用更充分的参数训练来达到更好的控制效果。为了实现对神经元脉冲发放规律作出快速、有效的估计,本文以真实脉冲样本的统计规律为依据,提出了一种考虑历史脉冲对发放强度影响的脉冲串概率模型--k-HS-λ模型。该模型的发放强度公式直观反映了真实脉冲样本的等待时间分布规律,其中包含两个未知参数c和a,参数c反映了无历史脉冲影响下的发放强度水平,参数a反映了历史脉冲对当前发放强度的影响。k-HS-λ模型参数的最佳值是通过对TRKS统计量的优化来确定的。最后,针对猴子快速脉冲神经元在接受感受野内刺激时发放的脉冲信号,分别使用标准泊松过程模型、k-HS-λ模型及gamma-TRRP模型进行建模实验。通过比较三种模型反映原始数据统计特性的准确程度,证明了k-HS-λ模型的实用性。
关键词: 脉冲串概率模型 k-HS-λ 脉冲发放强度 概率密度 KS图
For information in English, please click here
A simple and effective probabilistic model of spiking train
Abstract:The accurate estimation of the neuron spiking train is the basis for the spiking neuron networksv to simulate the biological intelligence more effectively. The modeling of spiking train has long been achieved by introducing more control factors and applying more sufficient parameter training. In order to make a fast and effective estimation of neuron spiking emission law, based on the statistical law of real pulse samples, a pulse train probability model, k-HS-λ model, which considers the influence of historical spiking train on the spiking intensity is proposed. The intensity formula of this model directly reflects the distribution law of the waiting time of real spiking train, including two parameters c and a. Parameter c reflects the intensity level without the influence of the historical spiking train, while parameter a reflects the influence of the historical spiking train on the current intensity. The optimal value for the k-HS-λ lambda model parameter is determined by optimizing the TRKS statistics. Finally, the standard Poisson process model, the k-HS-λ model and the gamma-TRRP model were respectively used to conduct modeling experiments for the pulse signals emitted by the fast spiking neuron when receiving the sensory field stimulation. The practicability of the k-HS-λ model is proved by comparing the accuracy of the three models in reflecting the statistical properties of the original data.
Keywords: probabilistic model of spiking train k-HS-λ intensity of spiking train probability density KS diagram
基金:
引用
No.****
同行评议
勘误表
一种简单有效的脉冲串概率模型
评论
全部评论0/1000