论文部分内容阅读
针对正则化极端学习机的隐层具有随机选择的特性,提出了一种增加删除机制来自适应地确定正则化极端学习机的隐层节点数.这种机制以对优化目标函数影响的大小作为评价隐层节点优劣的标准,从而淘汰那些比较“差”的节点,将那些比较“优”的节点保留下来,起到一个优化正则化极端学习机隐层节点数的目的.与已有的只具有增加隐层节点数的机制相比较,本文提出的增加删除机制在减少正则化极端学习机隐层节点数、增强其泛化性能、提高其实时性等方面具有一定的优势.典型混沌时间序列的实例证明了具有增加删除机制的正则化极端学习机的有效性和可行性.
Aiming at the hidden layer of regularization extreme learning machine, the hidden layer has the characteristics of random selection, and an adding deletion mechanism is proposed to adaptively determine the number of hidden nodes in the regularization extreme learning machine. The mechanism regards the influence on the optimization objective function as the evaluation Hidden layer nodes, so as to eliminate those nodes that are more “bad” and keep those nodes that are more “excellent”, so as to optimize the regularization of the number of hidden nodes in the extreme learning machine. Compared with the existing mechanisms that only increase the number of nodes in hidden layer, the proposed deletion and deletion mechanism has some advantages in reducing the number of hidden nodes in regularized extreme learning machine, enhancing its generalization performance and improving its real-time performance. An example of a typical chaotic time series proves the validity and feasibility of a regularized extreme learning machine with an added deletion mechanism.