摘要
本文研究了基于贝叶斯理论的神经网络算法,采用贝叶斯方法来确定超参数,使得神经网络在训练过程中能自适应地调节超参数的大小,得出目标函数的最优化参数,从而达到提高神经网络泛化能力的目的。还编制仿真软件,验证了该算法的可行性。
The neural network algorithm based on Bayes theory was researched in this paper. Using the Bayes method to confirm the parameters, the number of the parameters could be adjusted by the neural network, so as to get the optimization parameters of the objective function. So the generalization ability of the neural network was enhanced. The algorithm was proved to be feasible by compiling the simulation software.
出处
《光机电信息》
2011年第1期28-32,共5页
OME Information
关键词
神经网络
贝叶斯
正则化
neural network
Bayes
regularization
