BP网络过拟合满足的不确定关系式

Uncertainty Relation Satisfied With Overfitting of BP Neural Network

为了研究BP网络的过拟合现象 ,通过类比信息传递过程中的一般测不准关系式 ,建立了BP网络出现过拟合现象时的网络学习能力与推广能力之间满足的一般不确定关系式。通过模拟多种不同类型复杂程度函数的数值试验 ,确定出不确定关系式中的过拟合参数的取值范围为 1× 10 -5～ 5× 10 -4。根据一般不确定关系式 ,给出应用BP网络对给定样本集的训练过程中 ,避免出现过拟合现象 ,提高网络推广能力的方法。

In order to study the overfitting of BP neural network, a general uncertainty relation between the change of weighted value representing learning ability and the discrimination error of untraining sample sets representing generalization ability is revealed in the modeling of BP neural network by the analogy of general uncertainty relation in information transmitting process. Tests of numerical simulation for various of complicated functions are carried out to determine the value distribution (1×10\{-5\}～5×10\{-4\})of overfitting parameter in the uncertainty relation. Based on the uncertainty relation, the overfitting in the training process of given sample sets using BP network can be avoided.\;