Multi-objective differential evolution based on normalization and improved mutation strategy

Developing efficient algorithms for solving multi-objective optimization problems is a challenging and essential task in many applications. This task involves two or more conflicting objectives that need to be simultaneously optimized. Many real-world problems fall into this category. We introduce a...

Full description

Saved in:
Bibliographic Details
Published in:Natural computing 2017-12, Vol.16 (4), p.661-675
Main Authors: Awad, Noor H., Ali, Mostafa Z., Duwairi, Rehab M.
Format: Article
Language:English
Subjects:
Artificial Intelligence
Complex Systems
Computer Science
Evolutionary algorithms
Evolutionary Biology
Mathematical programming
Multiple objective analysis
Objectives
Optimization algorithms
Processor Architectures
Strategy
Theory of Computation
Citations: Items that this one cites
Items that cite this one
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Developing efficient algorithms for solving multi-objective optimization problems is a challenging and essential task in many applications. This task involves two or more conflicting objectives that need to be simultaneously optimized. Many real-world problems fall into this category. We introduce an improved version of multi-objective differential evolution (DE) algorithm, namely MO n DE that uses a new mutation strategy and a normalization method to select non-dominated solutions. The new mutation strategy “DE/rand-to- n best” uses the best normalized individual in terms of all the objectives to guide the search towards the true pareto optimal solutions. As a result, the probability of producing superior solutions is increased and a faster convergence is achieved. Summation of normalized objective values method is used instead of non-domination sorting to overcome the high computational complexity and overhead problems of sorting non-dominated solutions. The performance of our approach is tested on a set of benchmark problems that consist of two to five objectives. Different combinations of multi-objective evolutionary programming and multi-objective differential evolution algorithms have been used for comparisons. The results affirm the efficiency and robustness of the proposed approach among other well-known algorithms from the literature.
ISSN:1567-7818
1572-9796
DOI:10.1007/s11047-016-9585-y