Scholarlink

<p>Extreme Learning Machine (ELM), as a learning algorithm designed for single-layer neural networks, is much faster than conventional gradient-based learning algorithms. However, one of the main challenges of ELM lies in the random initialization of input layer weights, which can negatively impact its generalization performance and often requires a large number of hidden neurons. To address this limitation, this paper proposes an improved Particle Swarm Optimization (PSO) algorithm enhanced by a novel “CHAOS” operator inspired by natural and social bird behaviors. This operator increases the exploration capability of PSO, helping it to escape premature convergence and better explore complex and multimodal search spaces. The resulting CHAOS-PSO based ELM (CH-PSO-ELM) optimizes the weights of ELM to minimize computational error. The proposed method is evaluated on 18 benchmark datasets, including 12 UCI datasets and 6 medical datasets. The average improvements of CH-PSO-ELM compared to standard ELM, ELM-Differential Evolution (ELM-DE), ELM-Artificial Bees Colony (ELM-ABC), and ELM-PSO are 14.44%, 16.53%, 3.12%, and 2.12%, respectively.</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

Increasing the accuracy of an extreme learning machine using a chaos particle swarm optimization algorithm

  • 期刊论文

  • 发表日期：

    2025年11月27日

查看全文
  • Ghulam Rabani Romal Rasa,
  • Sepehr Ebrahimi Mood,
  • Elham Abbasi

摘要

Extreme Learning Machine (ELM), as a learning algorithm designed for single-layer neural networks, is much faster than conventional gradient-based learning algorithms. However, one of the main challenges of ELM lies in the random initialization of input layer weights, which can negatively impact its generalization performance and often requires a large number of hidden neurons. To address this limitation, this paper proposes an improved Particle Swarm Optimization (PSO) algorithm enhanced by a novel “CHAOS” operator inspired by natural and social bird behaviors. This operator increases the exploration capability of PSO, helping it to escape premature convergence and better explore complex and multimodal search spaces. The resulting CHAOS-PSO based ELM (CH-PSO-ELM) optimizes the weights of ELM to minimize computational error. The proposed method is evaluated on 18 benchmark datasets, including 12 UCI datasets and 6 medical datasets. The average improvements of CH-PSO-ELM compared to standard ELM, ELM-Differential Evolution (ELM-DE), ELM-Artificial Bees Colony (ELM-ABC), and ELM-PSO are 14.44%, 16.53%, 3.12%, and 2.12%, respectively.