摘要
本文对非监督学习方法和前馈神经网络的结构进行一些简单的改进。用改进后的结构和学习方法初始化前馈神经网络的权值 ,使前馈神经网络的训练速度和逼近精度显著提高。最后本文通过仿真实例验证了该方法。
In this paper,a little simple improvement was made on unsupervised learning principle and feedforward neural networks configuration.The synaptic weights of the networks is initiated with them.It is advanced notability for the networks learning.At last, the simulation is given to show the properties of the method.
出处
《微机发展》
2000年第3期1-3,共3页
Microcomputer Development
关键词
前馈神经网络
非监督学习方法
BP算法
Neural Networks
Unsupervised Learning
Initialization of the Synaptic Weights
Convergent Speed
Approximative Precision
