自适应神经网络学习方法研究

本文从连接权值、网络的拓扑结构、网络的学习参数以及神经元的激活特性等不同方面分别讨论了人工神经网络的学习问题，并就当前流行的ＢＰ模型提出了具体实现方法。实验表明，这些方法对于加快网络的收敛速度，优化网络的拓扑结构等方面有着显著成效，本文所述内容为ＡＮＮ学习算法的改进与设计提供了示例，途径和思想总结。

Adaptive Learning Method Study of Neural Networks

In this paper, the learning issue for artificial neural network is studied in several different aspects such as connection weight, network architecture, learning parameters and active function of unit. Some adaptive algorithms are proposed for the back propagation learning (BP), which is well known in the artificial neural network literature. The experiments show that they have good effect on increasing learning speed and training network architecture.