A novel chaotic teaching learning based optimization algorithm and its application in optimization of extreme learning machine. Issue 1 (August 2021)

• Record Type: Journal Article
• Authors: Ma, Yunpeng; Tang, Haoheng; Wang, Heqi; Wang, Zhenying; Zhang, Xinxin; Li, Lipeng
• Is Part Of: Journal of physics. Volume 2003:Issue 1(2021)
• Publication Date: 2021-08
• Subjects: Optimization -- Model -- Teaching Learning Based Optimization Algorithm -- Chaos Theory -- Extreme Learning Machine; Physics -- Congresses; 530.5
• DOI: 10.1088/1742-6596/2003/1/012003
• Languages: English
• ISSNs: 1742-6588

Abstract

Abstract: Recent ten years, the teaching learning based optimization algorithm (TLBO) has been widely concerned and successfully applied to solve various constraints and non-constraints problems. However, its convergence accuracy and convergence speed should be further improved. Therefore, a novel chaotic teaching learning based optimization algorithm (called CTLBO) is proposed. Firstly, chaotic variables are applied to initialize population individuals for increasing the diversity of population. Secondly, a kind of self-adaptive acceleration coefficient is introduced into teaching phase to enhance the convergence speed and solution quality. Finally, two population updating mechanisms are proposed to balance the exploration and exploitation capabilities in the learning phase. One is neighbor elitist search mechanism, another is chaos optimization mechanism. The performance of CTLBO is compared with five state-of-the-art optimization algorithms by several CEC mathematical problems. The experiment results show that the CTLBO yields better convergence rate than other algorithms on most testing functions. Additionally, the proposed CTLBO is applied to optimize the model parameters of extreme learning machine(ELM) and the tuned ELM is adopted to establish the NOx emissions model. Experiment results reveal that the NOx emissions model has good accuracy and meets the engineering requirement.