论文部分内容阅读
神经网络的学习能力与效率问题是神经网络研究的一个重要方向,该文基于正交变换提出一种网络正交学习算法,它具有学习速度快且能获得全局最优解的特点,并可有效地对学习过程中出现的异常情况进行求解,因而具有良好的普适性。同时对新样本的学习可在以前学习的基础上继续,使网络的学习具有循序渐进的特征,提高了学习效率。
The learning ability and efficiency of neural networks are an important direction of neural network research. Based on orthogonal transformation, this paper proposes a network orthogonal learning algorithm, which has the characteristics of fast learning speed and global optimal solution, and is effective To solve the abnormal situation in the learning process, it has good universality. At the same time, the learning of new samples can continue on the basis of previous learning, so that the learning of the network has the characteristics of gradual and gradual improvement and the learning efficiency is improved.