神经网络极速学习方法研究进展  

Survey of Extreme Learning of Neural Networks

作　　者：杨凤芝[1] 皮惠[1] 苏佳伟[2] YANG Feng-zhi, PI Hui, SU Jia-wei （1.School of Information, Yunnan Normal University, Kunming 650031, China; 2.School of Physics and Electronic Information, Yunnan Normal University, Kunming 650031, China）

机构地区：[1]云南师范大学信息学院,云南昆明650031 [2]云南师范大学物理与电子信息学院,云南昆明650031

出　　处：《电脑知识与技术》2011年第4期2368-2371,2474,共5页Computer Knowledge and Technology

摘　　要：神经网络已经在模式识别、自动控制及数据挖掘等领域取得了广泛的应用，但学习方法的速度不能满足实际需求。传统的误差反向传播方法（BP）主要是基于梯度下降的思想，需要多次迭代；网络的所有参数都需要在训练过程中迭代确定，因此算法的计算量和搜索空间很大。ELM（Extreme Learning Machine，ELM）是一次学习思想使得学习速度提高很多，避免了多次迭代和局部最小值．具有良好的泛化性能、鲁棒性与可控性。但对于不同的数据集和不同的应用领域，无论ELM是用于数据分类或是回归，ELM算法本身还是存在问题，所以本文对已有方法深入对比分析，并指出极速学习方法未来的发展方向。Neural Network have been widely applied in many fields including pattern recognition, automatic control, data mining etc. However, the traditional learning methods can not meet the actual needs. The traditional method is mainly based on gradient descent and it needs multiple iterations; all of the network parameters need to be determined by iteration. Therefore, the computational complexity and searching space will increase dramatically. ELM is one-time learning idea, this method is faster algorithm and voids a number of iterations and the local minimum, it has better generalization, robustness and controllability. But for different data sets and different applications, it is used for both data classification or regression. ELM algorithm has some problems. So this paper follow a comprehensive comparison and analysis of existing methods, future research directions are highlighted.

关 键 词：数据挖掘 神经网络 极速学习机 

分 类 号：TP18[自动化与计算机技术—控制科学与工程;自动化与计算机技术—控制理论与控制工程]