贝叶斯改进BP神经网络在织物染色配色中的应用

Research on Application for Color Matching in Textile Dyeing Based on BP Neural Network Adapted by Bayesian Regularization Algorithm

针对BP算法及其改进算法泛化能力不强的问题,探讨了用贝叶斯正规化算法与LM算法的结合来提高BP神经网络的泛化能力。结果表明,在相同网络规模或误差条件下,贝叶斯正规化算法泛化能力明显优于基本BP算法及其它改进的BP算法,且收敛速度较快。因此文中把贝叶斯正规化算法与LM算法结合应用到了织物染色的计算机配色中,其预测的配方和实验的数据比较接近,证明了该方法的可行性。

In order to solve the weak generalization capacity problem about the general and improved BP algorithms,how to raise generalization capacity of BP neural network by joining the Bayesian Regularization algorithm and the LM algorithm has been researched in this paper.Based on the same network size or error probability,the results show that Bayesian Regularization algorithm has better generalization capacity than the general and other improved BP algorithms,and it has higher convergent speed.The recipe of the method adapted by joining the Bayesian Regularization algorithm and the LM algorithm show that it is very close to the experiment data.This proves that the algorithm researched in this paper is feasible.