计算机科学
可扩展性
知识管理
人工智能
背景(考古学)
有可能
学习迁移
知识转移
能力(人力资源)
数据科学
心理学
古生物学
社会心理学
数据库
心理治疗师
生物
作者
Jiajun Zhou,Yun Tian,Liang Gao,Chao Lu,Xifan Yao
标识
DOI:10.1016/j.aei.2023.102343
摘要
Manufacturing services collaboration (MSC) is crucial to the industrial internet platform in determining the proper integration of multiple functionality unique services for a complex manufacturing process. As a critical enabler for resolving MSC, evolutionary algorithm (EA) plays an essential role in enhancing the efficiency of MSC optimization. However, EA solvers are generally executed from scratch and suffer from a high computational burden. Inspired by transfer learning, researchers have considered performing knowledge extraction across distinct problem instances to promote the problem-solving efficiency, giving rise to the evolutionary multi-task optimization (EMTO) paradigm. The appearance of EMTO brings an emerging knowledge-aware search paradigm that supports the online learning and exploitation of optimization experiences during the course of evolution process, thereby accelerating the search efficiency. In spite of competence in continuous problems, EMTO has received little visibility in the field of combinatorial optimization, particularly, MSC problem, let alone an experimental comparison of state-of-the-art EMTO approaches in the context of MSC. To fill in this void, this article explores the suitability of versatile EMTO solvers for addressing MSC and provides insights into the behavior of various knowledge transfer techniques in multi-task MSC environments. To our knowledge, this study is the first attempt that investigates the scalability of EMTO on MSC problems and the systematic evaluation of their performances. Our results unveil the intrinsic characteristics of distinct transfer techniques to MSC in different scenarios, along with a deep analysis of the abilities for EMTO alternatives on resolving MSC.
科研通智能强力驱动
Strongly Powered by AbleSci AI