Dual transfer learning with generative filtering model for multiobjective multitasking optimization

Multiobjective multitasking optimization (MTO) has attracted more and more attention because of its ability to solve multiple multiobjective optimization problems simultaneously. By transferring knowledge between tasks, MTO can improve the performance of optimization tasks. However, if the way of kn...

Full description

Saved in:
Bibliographic Details
Published inMemetic computing Vol. 15; no. 1; pp. 3 - 29
Main Authors Dang, Qianlong, Gao, Weifeng, Gong, Maoguo
Format Journal Article
LanguageEnglish
Published Berlin/Heidelberg Springer Berlin Heidelberg 01.03.2023
Springer Nature B.V
Subjects
Applications of Mathematics
Artificial Intelligence
Bioinformatics
Complex Systems
Control
Engineering
Evolutionary algorithms
Evolutionary computation
Filtration
Knowledge management
Mathematical and Computational Engineering
Mechatronics
Multiple objective analysis
Multitasking
Optimization
Performance enhancement
Regular Research Paper
Robotics
Evolutionary multitasking (EMT)
Dual transfer learning
Multiobjective optimization
Generative filtering model
Online AccessGet full text

Cover

More Information
Summary:Multiobjective multitasking optimization (MTO) has attracted more and more attention because of its ability to solve multiple multiobjective optimization problems simultaneously. By transferring knowledge between tasks, MTO can improve the performance of optimization tasks. However, if the way of knowledge transfer is not reasonable, it will have a negative impact on the performance of tasks. To solve this problem and ensure the effectiveness of knowledge transfer, this paper proposes a multiobjective evolutionary multitasking algorithm based on dual transfer learning with generative filtering model namely EMT–DLGM. Specifically, a dual transfer learning mechanism is proposed to reduce the difference between tasks and improve the efficiency of knowledge transfer through the global and local transfer strategies. Moreover, the generative filtering model is designed to generate promising solutions according to the multiple differential evolution operations and filtering model. The experimental results on three MTO test suites demonstrate that EMT–DLGM is superior or comparable to other state-of-the-art multiobjective evolutionary multitasking algorithms.
ISSN:1865-9284
1865-9292
DOI:10.1007/s12293-022-00374-9