Scholarlink

<p>Recently, the multilayer extreme learning machines (MLELMs) have been widely used to solve the practical problems, especially for the highly complicated and nonlinear cases. In this paper, we propose a novel kernel based multilayer online sequential extreme learning machine by using the multi-innovation identification theory. First, a sliding window of dynamical size based on the multi-innovation identification theory is introduced to the online sequential ELM autoencoders (OSELM-AEs) and a multi-innovation OSELM-AEs (MIOSELM-AEs) is given. Then, the multilayer MIOSELM-AEs are stacked to create a deep learning model (MLMIOSELM) for extracting the deeper feature from the multi-innovation of the previous layer. Finally, the kernel version of MLMIOSELM, named MLMIKOSELM, is derived to maintain the learning performance of dynamical nonlinearity. Simulation experiments are designed to test the effectiveness of the presented methods.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

A novel kernel based multilayer extreme learning machines based on multi-innovation identification theory

  • 期刊论文

  • 发表日期：

    2026年5月16日

查看全文
  • Yanjiao Wang,
  • Yiting Liu,
  • Weidi Li,
  • Muqing Deng

摘要

Recently, the multilayer extreme learning machines (MLELMs) have been widely used to solve the practical problems, especially for the highly complicated and nonlinear cases. In this paper, we propose a novel kernel based multilayer online sequential extreme learning machine by using the multi-innovation identification theory. First, a sliding window of dynamical size based on the multi-innovation identification theory is introduced to the online sequential ELM autoencoders (OSELM-AEs) and a multi-innovation OSELM-AEs (MIOSELM-AEs) is given. Then, the multilayer MIOSELM-AEs are stacked to create a deep learning model (MLMIOSELM) for extracting the deeper feature from the multi-innovation of the previous layer. Finally, the kernel version of MLMIOSELM, named MLMIKOSELM, is derived to maintain the learning performance of dynamical nonlinearity. Simulation experiments are designed to test the effectiveness of the presented methods.