Global optimization with one-class classification-assisted selection
Section snippets
Introduction and motivation
Evolutionary algorithms (EAs) [1] are a kind of heuristic optimization method [2]. EAs have attracted much attention for solving optimization problems [3] due to their advantage of achieving global optimum and almost-all-of-the-best-optimal-results. The main framework of an EA always includes three main components: (1) initialization, which initializes the original population of an optimization problem; (2) reproduction, which generates new trial solutions from the current population; and (3)
Related work
To save the number of FEs of EAs, especially when the FE is computationally expensive, researchers turn to other methods to assist in selection [13]. The widely used ones are surrogate (meta) model-based approaches (SAEAs) [13], [14], [15], [16], [17]. These algorithms build the models to approximate the original optimization problems. Then the model’s predicted estimated fitness values of solutions are used for selection instead of the true values. In this manner, SAEAs are able to reduce the
Our proposed one-class classification-assisted selection strategy
We formally define and formulate the optimization problem in this paper as follows:where is an n-dimensional decision variable vector; is the feasible region of the search space, and is the objective function.
As introduced in Section 1, a variety of EAs have been proposed to solve the optimization problem in Eq. (1). However, the efficiency of most EAs is still up for debate because there is a large number of FEs during the search for optimal results. We,
Experimental study
This section studies the performance of the proposed OCAS-EA approach. We first provide the experimental settings in Section 4.1. In Section 4.2, we compare the performance of OCAS-assisted EDA/LS and CoDE algorithms (denoted as OCAS-EDA/LS, OCAS-CoDE), with the original EDA/LS and CoDE. Section 4.3 studies the sensitivity of the OCAS strategy to the size of training dataset on OCAS-EDA/LS. Section 4.4 studies the sensitivity of the OCAS strategy to the kernel of SVM on OCAS-EDA/LS. Section 4.5
Conclusion and future work
We propose a one-class classification-assisted selection (OCAS) strategy for improving the performance of EAs. In the illustration of OCAS, first, the current population is defined as the positive training dataset. Next, the defined positive training dataset is used to build a one-class classifier. Then, the built classifier is used to predict the quality of the newly generated offspring solutions. Since the predicted unpromising solutions are thrown away before the fitness evaluation, the
CRediT authorship contribution statement
Jinyuan Zhang: Conceptualization, Data curation, Formal analysis, Investigation, Writing - original draft. Jimmy Xiangji Huang: Conceptualization, Formal analysis, Investigation, Writing - review & editing. Qinmin Vivian Hu: Conceptualization, Formal analysis, Investigation, Writing - review & editing.
Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
Acknowledgement
We gratefully appreciate the associate editor and all the four reviewers for their excellent comments that greatly helped to improve the quality of the article. This research is supported by the Natural Sciences and Engineering Research Council (NSERC) of Canada, the York Research Chairs (YRC) program and an ORF-RE (Ontario Research Fund-Research Excellence) award in BRAIN Alliance. All the work is done when the first author is a postdoc fellow sponsored by an NSERC discovery grant.
References (52)
- et al.
A review of concurrent optimisation methods
Int. J. Bio-Inspired Comput.
(2014) - et al.
Drift analysis and average time complexity of evolutionary algorithms
Artif. Intell.
(2001) - et al.
Towards an analytic framework for analysing the computation time of evolutionary algorithms
Artif. Intell.
(2003) Surrogate-assisted evolutionary computation: recent advances and future challenges
Swarm Evol. Comput.
(2011)- et al.
Evolutionary optimization with hierarchical surrogates
Swarm Evol. Comput.
(2019) - et al.
Preselection via one-class classification for evolutionary optimization
Proceedings of Genetic and Evolutionary Computation Conference Companion (GECCO 2018)
(2018) - et al.
Boosting evolutionary optimization via fuzzy-classification-assisted selection
Inf. Sci.
(2020) - et al.
Recent advances in differential evolution - an updated survey
Swarm Evol. Comput.
(2016) - et al.
Recent trends in the use of statistical tests for comparing swarm and evolutionary computing algorithms: practical guidelines and a critical review
Swarm Evol. Comput.
(2020) - et al.
Multiobjective evolutionary algorithms: a survey of the state of the art
Swarm Evol. Comput.
(2011)
Evolution strategies for continuous optimization: asurvey of the state-of-the-art
Swarm Evol. Comput.
Mining network data for intrusion detection through combining SVMS with ant colony networks
Future Gener. Comput. Syst.
Handbook of Evolutionary Computation
Evolutionary Optimization Algorithms
A survey on handling computationally expensive multiobjective optimization problems using surrogates: non-nature inspired methods
Struct. Multidiscip. Optim.
One-class classification: taxonomy of study and review of techniques
Knowl. Eng. Rev.
Evolutionary programming made faster
IEEE Trans. Evol. Comput.
A Gaussian process surrogate model assisted evolutionary algorithm for medium scale expensive optimization problems
IEEE Trans. Evol. Comput.
Problem Definitions and Evaluation Criteria for the CEC 2015 Competition on Learning-based Real-parameter Single Objective Optimization
Technical Report201411A, Computational Intelligence Laboratory, Zhengzhou University, Zhengzhou China and Technical Report
An estimation of distribution algorithm with cheap and expensive local search methods
IEEE Trans. Evol. Comput.
Differential evolution with composite trial vector generation strategies and control parameters
IEEE Trans. Evol. Comput.
Model-based evolutionary algorithms: a short survey
Complex Intell. Syst.
A comprehensive survey of fitness approximation in evolutionary computation
Soft Comput.
A review of surrogate assisted multiobjective evolutionary algorithms
Comput. Intell. Neurosci.
Evolutionary computations based on Bayesian classifiers
Int. J. Appl. Math.Comput. Sci.
Combining Bayesian classifiers and estimation of distribution algorithms for optimization in continuous domains
Connect. Sci.
Cited by (2)
A survey on state of art approaches in handling imbalance, positive and unlabelled data
2022, 3rd International Conference on Power, Energy, Control and Transmission Systems, ICPECTS 2022 - Proceedings