School of Automation and Electrical Engineering, Shenyang Ligong University, Shenyang, 110159, China.
Sci Rep. 2022 Aug 17;12(1):13938. doi: 10.1038/s41598-022-18267-9.
Biophysical models contain a large number of parameters, while the spiking characteristics of neurons are related to a few key parameters. For thalamic neurons, relay reliability is an important characteristic that affects Parkinson's state. This paper proposes a method to fit key parameters of the model based on the spiking characteristics of neurons, and improves the traditional particle swarm optimization algorithm. That is, a nonlinear concave function and a Logistic chaotic mapping are combined to adjust the inertia weight of particles to avoid the particle falling into a local optimum in the search process or appearing premature convergence. In this paper, three parameters that play an important role in Parkinson's state of the thalamic cell model are selected and fitted by the improved particle swarm optimization algorithm. Using the fitted parameters to reconstruct the neuron model can predict the spiking trajectories well, which verifies the effectiveness of the fitting method. By comparing the fitting results with other particle swarm optimization algorithms, it is shown that the proposed particle swarm optimization algorithm can better avoid local optima and converge to the optimal values quickly.
生物物理模型包含大量参数，而神经元的尖峰特征与少数几个关键参数有关。对于丘脑神经元来说，中继可靠性是影响帕金森状态的一个重要特征。本文提出了一种基于神经元尖峰特征来拟合模型关键参数的方法，并对传统的粒子群优化算法进行了改进。即，将非线性凹函数和 Logistic 混沌映射相结合，调整粒子的惯性权重，以避免粒子在搜索过程中陷入局部最优或出现过早收敛。本文选取丘脑细胞模型中对帕金森状态起重要作用的三个参数，用改进的粒子群优化算法进行拟合。使用拟合的参数来重构神经元模型可以很好地预测尖峰轨迹，验证了拟合方法的有效性。通过与其他粒子群优化算法的拟合结果进行比较，表明所提出的粒子群优化算法能够更好地避免局部最优并快速收敛到最优值。
