Abstract

Gene Expression Programming (GEP), a variant of Genetic Programming (GP), is a well established technique for automatic generation of computer programs. Due to the flexible representation, GEP has long been concerned as a classification algorithm for various applications. Whereas, GEP cannot be extended to multi-classification directly, and thus is only capable of treating an M-classification task as M separate binary classifications without considering the inter-relationship among classes. Consequently, GEP-based multi-classifier may suffer from output conflict of various class labels, and the underlying conflict can probably lead to the degraded performance in multi-classification. This paper employs evolutionary multitasking optimization paradigm in an existing GEP-based multi-classification framework, so as to alleviate the output conflict of each separate binary GEP classifier. Therefore, several knowledge transfer strategies are implemented to enable the interation among the population of each separate binary task. Experimental results on 10 high-dimensional datasets indicate that knowledge transfer among separate binary classifiers can enhance multi-classification performance within the same computational budget.

Highlights

  • Classification is a fundamental and active research topic in data mining

  • Gene Expression Programming (GEP) methods with different variation operators are employed with corresponding knowledge transfer techniques to show the effectiveness and the limits of the Evolutionary Multitasking methods in multi-classification, based on an existing multi-classification framework designed for GEP

  • As a member of evolutionary algorithms, Genetic Programming (GP) generally considers each solution for optimization problem as an individual of the whole population, in which the evolution of the algorithm is driven by variation operators encompassing mutation operators, crossover operators, and selection operators (Poli et al, 2008) among the individuals, like most meta-heuristic algorithms

Read more

Summary

INTRODUCTION

Classification is a fundamental and active research topic in data mining. Various real-world applications involving medical diagnosis, image categorization, credit approval, and etc., are covered by classification techniques. Considering the issue of the curse of dimensionality, many evolutionary algorithms (EA) have been utilized to assist aforementioned machine learning methods to tackle high-dimensional datasets, including Artificial Bee Colony (ABC) (Hancer et al, 2018), Particle Swarm Optimization (PSO) (Xue et al, 2012; Tran et al, 2018), and Genetic Programming (GP) (Chen et al, 2017) To be specific, these population-based algorithms can evolve individuals with a fitness function with respect to the machine learning classifier, and can be conducted in either single-objective or multi-objective fashion. GEP methods with different variation operators are employed with corresponding knowledge transfer techniques to show the effectiveness and the limits of the Evolutionary Multitasking methods in multi-classification, based on an existing multi-classification framework designed for GEP.

GEP MULTI-CLASSIFICATION FRAMEWORK
GP and GEP
AccGEP for Multi-Classification
18: Termination
MULTIFACTORIAL EVOLUTIONARY ALGORITHM
PROPOSED ALGORITHM
Framework
Evolution Process
Knowledge Transfer
Further Discussion
Findings
CONCLUSION
DATA AVAILABILITY STATEMENT
Full Text
Published version (Free)

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call