摘要
结合实际中神经网络的稀疏耦合特性,基于等式约束的L1优化方法下,本文发展了一类描述神经网络不规则放电动力学模型中耦合矩阵的重构方法,并通过数值模拟验证了该方法的有效性。与先前的基于SVD分解的重构方法相比较,本文的方法能在更短的观测时间内达到更高的重构精度。
Considering the sparse coupling characteristic of real neural network, a method based on the L1 optimization method with equality constraints is developed to reconstruct the coupling matrix of the dynamic model of neural networks with irregular firing rate. The effectiveness of proposed method is verified by numerical simulation. Compared with the previous reconstruction method based on SVD decomposition, the proposed method can achieve higher reconstruction accuracy in a shorter observation time.
出处
《应用数学进展》
2020年第8期1246-1254,共9页
Advances in Applied Mathematics
关键词
神经网络
L1优化
网络重构
Neural Networks
L1 Optimization
Network Reconstruction
