论文部分内容阅读
将ICBP网络模型引入BP-SOM(self-organizing feature maps)网络体系结构,并构建了一个基于SOM模型的集成型网络ICBP-SOM.BP-SOM是Ton Weijters提出的一种学习算法,它的目标是克服BP网络在特定类型的学习样本中存在的知识推广性方面的严重缺陷.提出此集成型网络的动机是,利用BP-SOM良好的知识解释能力和ICBP网络良好的推广性和自适应性构造一个ICBP-SOM模型,它具有良好的知识表示能力和极具竞争力的推广性能.在6个基准数据集上的实验结果验证了这一集成型网络的可行性和有效性.
The ICBP network model is introduced into the network architecture of BP-SOM (self-organizing feature maps) and an integrated network ICBP-SOM is built based on SOM model. The B-SOM is a learning algorithm proposed by Ton Weijters. Its goal Is to overcome the BP network in a particular type of learning samples existing knowledge extension of the serious flaw.The motivation of this integrated network is to use BP-SOM good knowledge of interpretation and ICBP network good promotion and adaptability Construct an ICBP-SOM model, which has good knowledge representation ability and highly competitive promotion performance.The experimental results on 6 benchmark datasets verify the feasibility and effectiveness of this integrated network.