A coevolutionary artificial bee colony for training feedforword neural networks

A coevolutionary artificial bee colony (CoABC) trainer based on a hybrid encoding mode is proposed to optimize the network structure and connection weights of a single-hidden layer feedforward network (SLFN). In the proposed CoABC, an integrated population (or colony) with double subpopulations, one...

Full description

Saved in:
Bibliographic Details
Published in:Neural computing & applications Vol. 37; no. 27; pp. 22649 - 22666
Main Authors: Zhang, Li, Li, Hong, Gao, Weifeng
Format: Journal Article
Language:English
Published: London Springer London 01.09.2025
Springer Nature B.V
Subjects:
Algorithms
Artificial Intelligence
Bees
Classification
Coding
Computational Biology/Bioinformatics
Computational Science and Engineering
Computer Science
Data Mining and Knowledge Discovery
Evolution
Genetic algorithms
Heuristic methods
Image Processing and Computer Vision
Neural networks
Optimization
Probability and Statistics in Computer Science
S.I.: Hybrid Approaches to Nature-inspired Optimization Algorithms and Their Applications
Saturn
Special Issue on Hybrid Approaches to Nature-inspired Optimization Algorithms and Their Applications
Swarm intelligence
Trainers
Training
Data classification
Structure learning
Artificial bee colony
Coevolutionary algorithm
Parameter optimization
Feedforward neural network
ISSN:0941-0643, 1433-3058
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:A coevolutionary artificial bee colony (CoABC) trainer based on a hybrid encoding mode is proposed to optimize the network structure and connection weights of a single-hidden layer feedforward network (SLFN). In the proposed CoABC, an integrated population (or colony) with double subpopulations, one of which is responsible for evolution of the network structure encoded as a binary vector, and the other of which is in charge of evolution of the connection weights encoded as a real-number vector, is utilized to coordinate the evolution of two subpopulations. Two types of updating formulas for binary and continuous variables in employed bees phase and onlooker bees phase are developed to enhance the search capability of CoABC. The CoABC can self-organize the structure and weights of a SLFN. In the experiments, 22 benchmark classification datasets are employed to evaluate the proposed CoABC trainer. The results show that the CoABC based on the hybrid encoding mode can train the optimal SLFNs for classification tasks with average test accuracy of 85.12%. Also, the proposed CoABC trainer outperforms original ABC-based trainers and other compared metaheuristic trainers as well as gradient-based trainers. Compared with all other algorithms, the proposed algorithm ranks top 1 in average test accuracy.
Bibliography:ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 14
ISSN:0941-0643
1433-3058
DOI:10.1007/s00521-024-10910-y