位置：成果数据库 > 期刊 > 期刊详情页
The construction and approximation of feedforward neural network with hyperbolic tangent function
  • 期刊名称：Appli. Math. J. Chinese Univ
  • 时间：2015.6
  • 页码：151-162
  • 分类：TP391[自动化与计算机技术—计算机应用技术;自动化与计算机技术—计算机科学与技术]
  • 作者机构：[1]中国计量学院理学院,浙江杭州310018
  • 相关基金：国家自然科学基金资助项目（No.61272023,61101240）.
  • 相关项目：球面散乱数据建模的神经网络方法
作者： Chen Zhixiang|Cao Feilong|
关键词： 支持向量机, 局部性质, 移动最小二乘, 回归估计, support vector machine, local property, moving least square, regression estimation
中文摘要：

基于移动最小二乘逐点逼近思想,移动权被引入到最小二乘支持向量机的误差变量中,得到新算法的模型.此外,证明了用移动最小二乘支持向量机作函数估计与在特征空间中用移动最小二乘法得到的解是一致的,揭示了移动最小二乘支持向量机所选择的核函数相当于移动最小二乘法所选择基函数组.数值试验与实例进一步验证所提出方法的优越性.

英文摘要：

Based on the idea of the moving least square for point by point approximation, the model of new algorithm was established by introducing the moving weighted function into error variance of least square support vector machine. It was proved that the solution of regression by using moving least square support vector machine coincided with the solution obtained by using moving least square method in the feature space. It also proved that the choice of kernel function of moving least square support vector machine was equivalent to the choice of basis function of moving least square method. The superiority of the proposed method is further shown by the numerical and the practice examples.

同期刊论文项目
球面散乱数据建模的神经网络方法
期刊论文 40 会议论文 3 著作 1
基于球调和分析理论的信号稀疏表示与重构算法
期刊论文 10
一类分裂等式问题的算法研究及其应用
期刊论文 1
同项目期刊论文
The construction and approximation of a class of neural networks operators with ramp functgions
球面神经网络的构造与逼近
Error estimates of qusai-interpolation and its derivatives
Conceptual and numerical comparisons of swarm intelligence optimization algorithms
Convergence rate of semi-supervised gradient learning algorithms, International Journal of Wavelets
Spherical scattered data quasi-interpolation by Gaussian radial basis function
The equivalent theorem for Jackson-type operators on spherical cap
The properties of logistic function and applications to neural network approximation
The Spectrum of A Parallel Repairable System with Warm Standby
The learning rate of l2-coefficient regularized classification with strong loss
The Spectral Analysis of k Out of N: G Redundant System
On the Equivalences and Differences of Evolutionary Algorithms
某些Mercer 核矩阵的权范数
Approximation by network operators with logistic activation functions
Biogeography-based optimization in noisy environments
Learning algorithm of neural networks on spherical Cap
Approximation by spherical neural networks with sigmoidal functions
Approximation by spherical de la Vallee-Poussin mean operators
Scattered data quasi-interpolation on spheres
Hybrid biogeogrqphy-based evolutionary algorithms
The learning rate of lp-coefficient regularized Shannon sampling algorithm
Update-based evolution control:A new fitness approximation method for evolutionary algorithms
The properties of logistic function and applications to neural network approximation，
The learning rate of regularized regression based on reproducing kernel Banach spaces
A modified extreme learning machine with sigmoidal activation functions
球面混合插值的误差估计
An effective fireworks algorithm for warehouse-scheduling problem
he learning rate of vector-valued ranking with least square loss
Scattered data approximation by neural networks operators
The construction and approximation of some neural networks operators
某些Mercer核矩阵的权范数
l~P-系数正则化Shannon采样学习算法收敛速度(英文)
结构化压缩感知研究进展
一种基于调和随机权网络与曲波变换的图像分类方法
随机权网络稀疏正则化算法及其应用
基于分层方法的白细胞五分类算法
球面广义移动最小二乘的diffuse泛函逼近
高效的白细胞分割算法
An oracle inequality for regularized risk minimizers with strongly mixing observations
Approximation by semigroup of spherical operators
Approximating and learning by Lipschitz kernel on the sphere