BP神经网络的双重优化的改进研究

Research on the Improvement of Dual Optimization on BP Neural Network

目前,在交通、农业以及相关数据挖掘等领域应用最为广泛的BP网络模型是较为重要的一种神经网络算法模型,但其性能一直达不到理想状态。BP神经网络收敛性、预测精度一般,并且容易陷入局部最优解,这些缺点需不断改善。针对上述提到问题提出采用动态自学习影响因子和改进网络激活函数两者相结合的一种改进BP网络算法。实验表明,提出改进BP网络方案能够大幅度提升BP神经网络的收敛效率以及精度。

At present,the most widely used BP network model in traffic,agriculture and related data mining is still an important neural network algorithm model,but its performance has not been satisfactory.The convergence and prediction accuracy of the BP neural network are general and easy to fall into the local optimal solution,and these shortcomings still need to be improved continuously.Therefore,in view of the above mentioned problems,an improved BP network algorithm is proposed by combining the dynamic self-learning effect factor and the improved network activation function.Experiments show that the proposed improved BP network scheme can greatly improve the convergence efficiency and accuracy of BP neural networks.