Scribd
Upload
9 views1 page

Adaptive Differential Evolution With Intersect Mutation and Repaired Crossover Rate

Differential Evolution, categories itself as a non-linear, non-differentiable and stochastic direct search method over continuous space, aims at global optimization of objective function. A hybrid differential evolution technique is proposed with intersect mutation and repaired crossover rate for adaptive DE.

Uploaded by

itikhumlwng
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
9 views1 page

Adaptive Differential Evolution With Intersect Mutation and Repaired Crossover Rate

Differential Evolution, categories itself as a non-linear, non-differentiable and stochastic direct search method over continuous space, aims at global optimization of objective function. A hybrid differential evolution technique is proposed with intersect mutation and repaired crossover rate for adaptive DE.

Uploaded by

itikhumlwng
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 1

Proceedings of International Conference on Computational Intelligence & IoT (ICCIIoT) 2018

Adaptive differential evolution with intersect mutation and repaired


crossover rate

Tanusree Poddera,*,Diptendu Bhattachyab,Srimantini Chakrabortyb


a
Techno College of Engineering Agartala, Tripura, 799035, India
b
National Institute of Technology Agartala, Tripura, 799046, India

Abstract: Differential Evolution, categories itself as a non-linear, non-differentiable and stochastic direct search method over continuous space, aims at
global optimization of objective function. A hybrid differential evolution technique is proposed with intersect mutation and repaired crossover rate for
adaptive DE. The Mutation technique is applied on divided worst and better part of the population to direct the search mechanism. Repaired crossover rate
using binary string of successful crossover, which is an addition to JADE for better parameter adaptation phase also employed for better performance in
the proposed algorithm. The proposed approach compared with DE shows improved result for different benchmark function. In future studies, the method
will be compared with other DE based algorithm to examine the success rate.

An optimization problem with several conflicting objectives is broadly


1. Introduction known as Multi-objective Optimization. Zhou et al. proposed a survey on
different multi-objective algorithm which include Multi-objective
In the chronicle of Optimization, Evolutionary Algorithm (EAs) has Evolutionary Algorithm (MOEAs), Memetic MOEAs, Decomposition
gained most popularity for solving optimization problem. DE as a variant based MOEAs (MOEA/Ds) etc. The main focus of NSGA-II introduces
of EAs has shown promising result in solving real world problems. These by Deb et al. is on constrained multi-objective optimization, for which
days many algorithms exists for optimization using gradient and heuristic they proposed a fast non-dominated sorting approach combined with
based searched technique in deterministic and scholastic contexts. Usually crowding distance sorting to select the most appropriate non-dominated
to broaden the pertinence of evolutionary algorithms various physical and front for the ne et al. in implemented multi-objective
natural are mimicked and mathematically formulated. optimization with differential evolution (DEMO) and for selecting the
Storn, et. al, introduced DE for continuous domain optimization. DE is individual for next generation DEMO uses crowding distance and pareto
used and applied extensively due to its simple construction mechanism, based ranking method.
robust behavior, ease of use and convergence speed. For search
mechanism it uses mutation and selection to direct the search towards
potential feasible regions. Nomenclature
Zhang et al. proposed a new strategy "DE/current-to-pbest" and automatic
parameters adaption phase. Out of the top 100% individuals of the current NP total no of individual in population
population pbest value has been selected; F and Cr values are d denotes dimension of each individual or total no of elements present in
automatically updated based on success rate. By forming a binary string of each individual
average no. of successful donor vector, crossover rate is repaired to get a gmax maximum generation Structure
better trial vector proposed by Gong et al. By using the concept of work Xi
specialization, a novel adaptive multiple sub-population based DE
(MAPDE) algorithm is proposed by Cui et. al., in which by ordering the
individual of current population in non-decreasing order of objective
value sub-divided into three sub-groups, three dissimilar strategies are
used for these three sub-groups including a parameter adaptation phase for
efficaciously choosing F and Cr using Cauchy and Gaussian distribution
for preliminary generation and restructured subsequently using the success 2. Traditional Differential Evolution
rate of F and Cr.
Rakshit et al. developed a noise handling tool using Multi-objective Like other evolutionary algorithm DE performs three operations on each
Differential Evolution based scheme for periodic objective function population, which includes mutation, crossover and selection. The
evolution by adaptive selection of the sample size as well as technique to performance of DE relies mainly on the parameter settings of mutation
determine expected value for noisy objective sample, to get the trial factor F, crossover ratio Cr and population size NP. The value of F directs
solution by probabilistic selection induced crowding distance. the search mechanism, Cr directs the convergence speed and NP is
generally user defined.

HTTPS://WWW.SSRN.COM/LINK/IJCIIOT-PIP.HTML ELSEVIER-SSRN (ISSN: 1556-5068)

424

You might also like