论文部分内容阅读
本文以天然留兰香的组分构成与其品质的关系为例,讨论人工神经元方法用于复杂信息模式分类的问题。提出一种广义的误差反传训练策略,将网络的训练范围从联接权扩大到神经元模型。这种新的训练方法(GBP)能提高多层前传网络的学习效率,加快收敛的速率。实际运行的结果表明,所需训练时间仅为普通误差反传(BP)训练方法的1/15,并能达到较高的预报精度。
In this paper, the composition of natural Spearmint and the relationship between the quality of its as an example, the method of artificial neurons for the classification of complex information model issues. A generalized error backtracking training strategy is proposed, which expands the network training range from the connection right to the neuron model. This new training method (GBP) can improve the learning efficiency of multi-layer network and accelerate the rate of convergence. The actual operation results show that the required training time is only 1/15 of the normal error back propagation (BP) training method, and can achieve higher forecast accuracy.