节点文献
前向神经网络学习算法研究
Learning algorithm for feedforward neural networks
【摘要】 利用优化算法中的信赖域法,选取了合适的计算模型,并具有全局收敛性、收敛快速性。BP网络学习算法对异或问题、逼近问题进行的具有统计意义的仿真,表明了该算法的有效性及对样本数量变化不敏感的特性。
【Abstract】 :Presents a new learning algorithm developed using a trust region method for BP networks with overall convergence and enhanced convergence rate and concludes from simulation results that the new algorithm is effective.
【关键词】 BP网络;
阻尼最小二乘法;
前向神经网络;
误差反传;
信赖域法;
逼近问题;
ABPM算法;
【Key words】 BP network; Least square method with damper; feedforward neural networks; error back propagation; trust region method; approach problem; ABPM algorithm;
【Key words】 BP network; Least square method with damper; feedforward neural networks; error back propagation; trust region method; approach problem; ABPM algorithm;
- 【文献出处】 哈尔滨建筑大学学报 ,Journal of Harbin University of Civil Engineering and Architecture , 编辑部邮箱 ,2001年01期
- 【分类号】TP183
- 【被引频次】10
- 【下载频次】116