摘要
构建了一种基于IF模型的侧抑制神经网络群,用以实现位置定位.采用基于H-H模型简化的IF模型构造神经网络群并基于概率密度分布进行位置定位.在神经网络群学习过程中,运用PITS(progressive interactive training scheme)方法进行参数学习,利用信息中心(IC)储存每次训练的结果,在保证输出收敛的情况下,比较跟踪结果的误差函数给出权值调整公式进行自学习.实验结果表明:基于IF模型构建的神经网络群可以实现位置定位.采用H-H模型简化的IF模型提高了学习效率和定位速度;运用PITS算法进行参数学习提高了定位精度.
A kind of neural network ensembles based on the integrate fire ( IF) model is proposed to achieve the purpose of position fixing. The neural network ensembles are constructed by the IF model which is based on the optimized H-H model. The position fixing procedure is based on probability distribution in a motion environment. During the training procedure,the parameters are selflearned by the progressive interactive training scheme ( PITS) ,and the weights are adjusted to selfstudy by comparing the error function of the tracking result of each step under the condition of output convergence. Experimental results show that the proposed neural network ensembles can be used in the field of position fixing. The simplified IF model improves the learning efficiency and the positioning speed while the PITS learning algorithm increases the accuracy of position fixing.
出处
《东南大学学报(自然科学版)》
EI
CAS
CSCD
北大核心
2010年第S1期178-182,共5页
Journal of Southeast University:Natural Science Edition
关键词
侧抑制机制
神经网络群
定位
PITS算法
lateral inhibition
neural network ensembles
position fixing
progressive interactive training scheme ( PITS)