摘要
针对神经网络的结构存在冗余的问题,提出了一种利用粗糙集优化神经网络结构的方法。在保持神经网络处理能力的基础上,利用网络的隐层神经元与网络误差构造决策表并进行属性约简,删除冗余的隐层节点。实验证明,该方法可以简化神经网络结构和减少神经网络的训练时间。
In allusion to the redundancy of neural network structure,a optimizing method of neural network structure based on rough sets is established.Without changing the net＇s process capacity,this method builts a decision table by using hidden-layer neurons and error of the network,and then use attribute reduction,delete redundancy neurons in the hidden-layer.The experiment is made and prove that this mathod can facilitate the structure of the neural networks and reduce the training time.
出处
《计算机与数字工程》
2010年第5期49-51,共3页
Computer & Digital Engineering
关键词
神经网络
粗糙集
属性约简
网络结构
neural network
rough sets
attribute reduction
net structure
