论文部分内容阅读
ｎ维超立方体顶点的分类问题是人工神经网络研究中的重要问题之一。若对ｎ维超立方体的顶点进行正确分类，同时保证网络具有最好的稳健能力，则任两个不同类顶点连线的中点都应是分割这两顶点的超平面上的点。基于这样的思想，本文导出了使网络稳健能力最强的分类超平面的标准方程，给出了网络各层节点之间连接权值和阈值的可能值。其连接权值仅需取＋１、－１和０，阈值仅需取１２加上〔－ｎ，ｎ－１〕上的整数，从而可获得最优的网络结构、最少的隐节点数目、最大的稳健能力，这样结构的网络易于训练，并不易进入局部极小点。
The classification of n-dimensional hypercube vertices is one of the most important problems in the research of artificial neural networks. If the vertices of the n-dimensional hypercubes are correctly classified and the network has the best robustness, the midpoint between any two different vertex lines should be the point on the hyperplane that divides the two vertices. Based on this idea, this paper derives the standard equation of classification hyperplane that makes the network robust, and gives the possible values of the connection weights and thresholds between nodes in different layers of the network. The connection weights only need to take +1, -1 and 0, and the threshold only needs to add 12 plus the integer on [-n, n-1], so as to obtain the optimal network structure, the minimum number of hidden nodes and the maximum Robust ability, this structure of the network easy to train, and not easy to enter the local minimum.