基于小样本数据的BP神经网络建模

针对小样本条件下BP(back propogation)神经网络存在预测精度不高的问题,将专家知识融入BP神经网络训练过程中解决此问题.首先BP神经网络通过遗传算法获得最优初始权值和阈值;其次对专家知识进行数学表达;最后通过增广拉格朗日乘子法将专家知识融入BP神经网络训练过程中.利用实际中的结晶动力学问题对所提方法进行...

BP Neural Network Modeling Based on Small Sample Data

Under the condition of small samples, BP neural network training has the question of the prediction accuracy is not high, it is proposed to integrate expert knowledge into the training process of BP neural network. First, Obtain the optimal initial weight and threshold of BP neural network through genetic algorithm; secondly, Mathematical expression of expert knowledge; finally, The augmented Lagrange method is used to blend the expert knowledge with the training process of BP neural network . The method proposed in this paper is verified by practical crystallization kinetics issue. The simulation results show that, Compared to existing methods for solving small sample questions, the BP neural network combine with the expert knowledge under small sample conditions can effectively improve the accuracy of training consequence.