论文部分内容阅读
Spiking神经网络是以更具有生物性质的Spiking神经元为基本单元构成的第三代人工神经网络。Spiking神经网络学习算法的关键是构成合适的突触权值学习规则。考虑突触前后神经元产生的脉冲时间差对突触权值的影响，将梯度下降算法和脉冲时间依赖的可塑性（STDP）学习规则相结合，建立了基于STDP规则的误差反馈Spiking神经网络的突触权值学习算法。仿真实验对传统的BP算法、传统Spiking神经网络算法及改进的算法进行了比较，应用典型的数据集进行测试可以得出改进的算法在预测准确率和收敛速度方面都得到了提高。新的突触权值学习算法结构简单，对于非线性系统预测问题精度较高。
Spiking neural network is the third generation of artificial neural network composed of the more biological Spiking neurons as the basic unit. The key of Spiking neural network learning algorithm is to form suitable learning rules of synaptic weights. Considering the influence of the difference of pulse time between presynaptic neurons on the synaptic weight, the gradient descent algorithm and the pulse time-dependent plasticity (STDP) learning rule were combined to establish the synchrotron of error feedback Spiking neural network based on STDP rule Weight learning algorithm. Compared with the traditional BP algorithm and the traditional Spiking neural network algorithm and the improved algorithm, the simulation experiments show that the improved algorithm can improve the prediction accuracy and the convergence rate by using typical datasets. The new synaptic weight learning algorithm has a simple structure and high precision for nonlinear system prediction problems.