一种基于模糊推理的神经网络学习算法

提出了一种新的神经网络学习算法．相对于其他学习算法，该算法侧重于网络参数的调整，通过对样本集的模糊推理、调整和分类学习来实现自适应的神经网络学习．结果表明，该算法能大大提高神经网络的学习速度和学习效率，并能从样本集中得到反常样本和小概率事件样本，对小概率事件样本有很好的学习能力．

A Neural Network Learning Algorithm Based on Fuzzy Deduction

A new algorithm was proposed which improves the training of neural networks. Different from previous approaches, this new approach focuses on the samples. With self-adaptive fuzzy deduction, classification and classified training, it achieves a better training performance. This algorithm was tested and works quite well. Besides its efficiency, the good ability to detect abnormal and low-probable samples is also worth notice.