摘要
提出一种既能改变网络结构又能不断添加新数据的学习算法,克服了在线序列ELM一旦开始学习便无法调整网络结构的困难。基于在线序列ELM算法的基本思想和理论及目前面临的难点,结合已存在ELM及改进算法,探寻适合在线序列ELM算法的网络结构调整策略。误差渐减在线序列ELM算法可以解决在线序列ELM目前面临的挑战,且所提算法理论实验方面均切实可行。
In this paper, a learning algorithm that can not only transform network structure but also add to new data is proposed, it not only learns online data hut also self-adapts network meanwhile. Solv- ing the challenge that Online Sequence Extreme Learning Machine(OS-ELM) could not change struc- ture once the learning process starts. So we find a strategy to adjust the framework of network for OS-ELM which base on theory and some improved algorithms on OS-ELM and Extreme Learning Ma- chine(ELM). Error Decrease Online Sequence Extreme Learning Machine can deal with the difficulty that OS-ELM is facing, and the algorithm proposed is reasonable in theory and numerical value experi- ment.
出处
《咸阳师范学院学报》
2014年第2期5-8,共4页
Journal of Xianyang Normal University
关键词
在线序列
误差渐减
网络结构
online sequence
extreme learning machine
network