An Adaptive Archive-Based Evolutionary Framework for Many-Task Optimization

Chen, Yongliang; Zhong, Jinghui; Feng, Liang; Zhang, Mengjie

doi:10.1109/tetci.2019.2916051

Cited by 130 publications

(58 citation statements)

References 33 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…As illustrated in algorithm 3, the transfer process is launched whenever the iteration count t can be divided by a certain transfer interval δ. Unlike those methods that try to adaptively select a similar task to transfer (Chen et al, 2019), as a preliminary study, this paper simply randomly selects an archive A j for each subpopulation POP i , where i may not necessarily differ from j due to the discussion above.…”

Section: Evolution Processmentioning

confidence: 99%

“…The knowledge transfer has been investigated in various population-based algorithms, and the investigation mainly concentrated on the chromosome representation (Zhou et al, 2016;Zhong et al, 2018a), and the problem similarity (Da et al, 2018;Chen et al, 2019). However, in this paper, the problem representation for each binary classification problem does not require redesign, and we tend to select the archive randomly to assist the target task.…”

Section: Knowledge Transfermentioning

confidence: 99%

“…Equipped with the capability of latent genetic transfer, Evolutionary Multitasking can resolve many optimization problems simultaneously by enabling the knowledge transfer among different problems through the unified chromosome representation. In control of the synergies of searching space for varying optimization tasks (Gupta et al, 2016a,b;Da et al, 2018;Zhou et al, 2018), Evolutionary Multitasking, which can be easily employed on existing population-based algorithm (Feng et al, 2017;Chen et al, 2018;Liu et al, 2018;Zhong et al, 2019), have shown promising results on a vast number of cases in multi-objective optimization (Gupta et al, 2016c;Feng et al, 2018), symbolic regression (Zhong et al, 2018a), capacitated vehicle routing problems (Zhou et al, 2016), expensive optimization tasks (Min et al, 2017), and can be extended to a large scale version (Chen et al, 2019;Liaw and Ting, 2019) to enable some more scalable applications in the future. The methodology of Evolutionary Multitasking paradigm naturally fits the multi-classification problem, by treating each binary classification problem as an optimization task within certain function evaluations.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

A Preliminary Study of Knowledge Transfer in Multi-Classification Using Gene Expression Programming

Wei

Zhong

2020

Front. Neurosci.

Self Cite

View full text Add to dashboard Cite

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 interrelationship 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.

show abstract

Section: Evolution Processmentioning

confidence: 99%

Section: Knowledge Transfermentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

A Preliminary Study of Knowledge Transfer in Multi-Classification Using Gene Expression Programming

Wei

Zhong

2020

Front. Neurosci.

Self Cite

View full text Add to dashboard Cite

show abstract

“…Therefore, it is important to choose the most suitable task (or assisted task) to be paired with the present task (or target task) for knowledge transfer. An adaptive selection mechanism of choosing suitable task was proposed by simultaneously considering the similarity between tasks and the accumulated rewards of knowledge transfer during the evolution [38].…”

Section: Multi-population Evolution Modelmentioning

confidence: 99%

Rigorous Analysis of Multi-Factorial Evolutionary Algorithm as Multi-Population Evolution Model

Wang¹,

Xu²,

Rong³

et al. 2019

IJCIS

View full text Add to dashboard Cite

A B S T R A C TMulti-task optimization algorithm is an emergent paradigm which solves multiple self-contained tasks simultaneously. It is thought that multi-factorial evolutionary algorithm (MFEA) can be seen as a novel multi-population algorithm, wherein each population is represented independently and evolved for the selected task only. However, the theoretical and experimental evidence to this conclusion is not very convincing and especially, the coincidence relation between MFEA and multi-population evolution model is ambiguous and inaccurate. This paper aims to make an in-depth analysis of this relationship, and to provide more theoretical and experimental evidence to support the idea. In this paper, we clarify several key issues unsettled to date, and design a novel across-population crossover approach to avoid population drift. Then MFEA and its variation are reviewed carefully in view of multi-population evolution model, and the coincidence relation between them are concluded. MFEA is completely recoded along with this idea and tested on 25 multi-task optimization problems. Experimental results illustrate its rationality and superiority. Furthermore, we analyze the contribution of each population to algorithm performance, which can help us design more efficient multi-population algorithm for multi-task optimization. MTO paradigm [6]. If the optimization tasks happen to bear some commonality or complementarity, then the inclusion of knowledge transfer often leads to significant performance improvements relative to conventional EAs alone.

show abstract

“…Some preliminary work was carried out suggesting that MFEA is being able to capitalize the knowledge overlap across relevant tasks to facilitate better objective value on solving multiple combinatorial problems concurrently [11,22]. Other studies attempt to investigate the theoretical foundation of MFEA as well as improve this algorithm such that its performance would not be impeded by solving tasks that are two conflicting in their fitness landscape and global optima [2,5]. Also, there is no prior attempt to design a MFEA for solving CluMRCT.…”

Section: Related Workmentioning

confidence: 99%

Multifactorial Evolutionary Algorithm For Clustered Minimum Routing Cost Problem

Trung

Thanh

Hieu

et al. 2019

Proceedings of the Tenth International Symposium on Information and Communication Technology - SoICT 2019

View full text Add to dashboard Cite

Minimum Routing Cost Clustered Tree Problem (CluMRCT) is applied in various fields in both theory and application. Because the CluMRCT is NP-Hard, the approximate approaches are suitable to find the solution for this problem. Recently, Multifactorial Evolutionary Algorithm (MFEA) has emerged as one of the most efficient approximation algorithms to deal with many different kinds of problems. Therefore, this paper studies to apply MFEA for solving CluMRCT problems. In the proposed MFEA, we focus on crossover and mutation operators which create a valid solution of CluM-RCT problem in two levels: first level constructs spanning trees for graphs in clusters while the second level builds a spanning tree for connecting among clusters. To reduce the consuming resources, we will also introduce a new method of calculating the cost of CluMRCT solution. The proposed algorithm is experimented on numerous types of datasets. The experimental results demonstrate the effectiveness of the proposed algorithm, partially on large instances. [12].Among those mentioned problems, CluMRCT is one of the most newly investigated problems. This problem has been formally formulated in [12]. Concretely, consider a connected, undirected graph G = (V , E, w) with nonnegative edge length function w :A spanning tree T of G is a clustered spanning tree if T can be cut into k subtrees by eliminating k − 1 edges such that each subtree is a spanning tree for only one cluster. The CluMRCT problem focuses on the routing cost, which is the sum of shortest path distance between any pairs of vertices given a clustered spanning tree T . The CluMRCT problem is finding on graph G a clustered spanning tree T having the minimum routing cost. One practical realization of this problem is in computer network application where the communication terminals are vertices, and they are partitioned into many clusters. The communication between those terminals are restricted within a cluster and only a few terminals can be connected to another cluster for maximizing efficiency and other security concerns. Solving CluMRCT is equivalent to facilitate the network architecture that consumes the minimum peer-to-peer communication resources.

show abstract

An Adaptive Archive-Based Evolutionary Framework for Many-Task Optimization

Cited by 130 publications

References 33 publications

A Preliminary Study of Knowledge Transfer in Multi-Classification Using Gene Expression Programming

A Preliminary Study of Knowledge Transfer in Multi-Classification Using Gene Expression Programming

Rigorous Analysis of Multi-Factorial Evolutionary Algorithm as Multi-Population Evolution Model

Multifactorial Evolutionary Algorithm For Clustered Minimum Routing Cost Problem

Contact Info

Product

Resources

About