论文部分内容阅读
在判决神经网络（ＤＢＮＮ）的基础上提出了一种基于模糊算法的模糊判决神经网络（ＦＤＢＮＮ）．在网络训练中引入置信度和容噪度的概念，提高了网络分类的稳定性，同时克服了（ＤＢＮＮ）在训练样本混有噪声时学习困难和泛化能力不高的缺点．因ＦＤＢＮＮ在学习时的不均匀性，大大加快了网络训练的时间，提高了训练的效率．实验结果表明，ＦＤＢＮＮ的性能高于ＢＰ网，而且也比ＤＢＮＮ在稳定性和识别率上有了显著的提高．
Based on the decision neural network (DBNN), a fuzzy decision neural network (FDBNN) based on fuzzy algorithm is proposed. The concept of confidence and noise tolerance is introduced in network training, which improves the stability of network classification and overcomes the shortcomings of (DBNN) learning difficulty and poor generalization ability when mixed with training samples. Due to the inhomogeneity of FDBNN in learning, it greatly speeds up the network training time and improves the training efficiency. The experimental results show that the performance of FDBNN is higher than that of BP network, and it also improves the stability and recognition rate more than DBNN.