Negatively Correlated Search for Constrained Optimization

Yuan LI, Xiaofen LU, Xin YAO

Research output: Book Chapters | Papers in Conference ProceedingsConference paper (refereed)Researchpeer-review


Evolutionary algorithms (EAs) combined with constraint handling techniques (CHTs) are very effective in solving constrained optimization problems (COPs). Recently, negatively correlated search (NCS) has been shown to be powerful for real-world multimodal optimization problems (MMOP). However, there has been few works on utilizing NCS to solve COPs. In this paper, we present a novel constrained optimization evolutionary algorithm (COEA) named NCS-E which combined NCS with a new CHT ensemble method to deal with COPs. It integrates three complementary CHTs as voters, and individuals are first voted by each voter, then the individual with more weighted votes is considered better. In addition, the negative correlation information of the voted individuals is utilized to adaptively adjust the voting weights. To demonstrate the performance of NCS-E, NCS-E is compared with four NCS-based COEAs and two state-of-the-art methods, MAgES and MAgES-VMCH, on 57 real-world COPs. The statistical results show that NCS-E exhibits the best performance among the compared methods. Moreover, NCS-E was applied to solve the Deep Neural Network pruning problem and compared with the state-of-the-art pruning method OLMP on three reference models. Empirical results show that the proposed method improves the performance of OLMP on all three DNN pruning tasks. © 2023 IEEE.
Original languageEnglish
Title of host publication2023 IEEE Congress on Evolutionary Computation, CEC 2023
PublisherInstitute of Electrical and Electronics Engineers Inc.
Number of pages10
ISBN (Print)9798350314588
Publication statusPublished - 1 Jul 2023
Externally publishedYes
EventIEEE 2023 Congress on Evolutionary Computation - Chicago, United States
Duration: 1 Jul 20235 Jul 2023


ConferenceIEEE 2023 Congress on Evolutionary Computation
Abbreviated titleCEC 2023
Country/TerritoryUnited States
Internet address

Bibliographical note

This work was supported by the National Natural Science Foundation of China (Grant No. 61906082), Guangdong Provincial Key Laboratory (Grant No. 2020B121201001), the Program for Guangdong Introducing Innovative and Entrepreneurial Teams (Grant No. 2017ZT07X386), the Research Institute of Trustworthy Autonomous Systems (RITAS).


  • Constrained optimization
  • DNN pruning
  • Ensemble
  • Evolutionary algorithm
  • Negatively correlated search


Dive into the research topics of 'Negatively Correlated Search for Constrained Optimization'. Together they form a unique fingerprint.

Cite this