角分类前向神经网络研究

角分类算法是一类快速分类算法，以其为学习算法的前向神经网络，在信息检索，特别是在线信息检索等领域有着重要的应用．通过对CC4学习算法的分析，揭示了泛化距离在角分类神经网络中的意义．针对文本数据的快速分类要求，提出了新的角分类网络TextCC．为解决数据的多类别判定问题，给出了新的角分类神经网络隐层与输出层之间连接矩阵的学习算法．实验表明，新的角分类神经网络隐层与输出层之间连接矩阵的学习算法有效，TextCC的分类精度教CC4的分类精度显著的提高．

Research on Forward Neural Network for Corner Classification

Corner classification (CC) is a kind of algorithms for instantly classification. The forward neural network trained by CC algorithm has important role on information retrieval, especially online information retrieval. The training algorithm of CC4 is analyzed in this paper and generalized distance is used to interpret the principle of training algorithm for CC4. A new network for CC, named as TextCC, is presented for classifying text object instantly. To give a solution for multi-corner judging, new training algorithm for the construction of connection matrix between hidden and output layer of forward neural network trained is presented. The experimental results show that the new training algorithm can work well on corner classification task and the precision of TextCC is improved markedly than CC4's.