As we know, exploration and exploitation are two important criteria in a population-based metaheuristic. I could find some formula to obtain a quantitative comparison of the exploration ability among different algorithms. but how we can compare the exploitation ability quantitatively among different metaheuristic algorithms?
Dear Seyed,
I suggest you to see links in subject.
-Metaheuristic research: a comprehensive survey - ResearchGate
https://www.researchgate.net/.../322478953_Metaheuristic_researc...
-comparative analysis of methods for multi-objective structural ...
www.gf.uns.ac.rs/~zbornik/doc/NS2016.007.pdf
-Harmony Search as a Metaheuristic Algorithm - arXiv
https://arxiv.org/pdf/1003.1599
-A Comprehensive Survey on Particle Swarm Optimization Algorithm ...
https://www.hindawi.com/journals/mpe/2015/931256/
-A whale optimization algorithm (WOA) approach for clustering ...
https://www.cogentoa.com/article/10.../25742558.2018.1483565
-COMPARISON OF META-HEURISTIC ALGORITHMS FOR SOLVING ...
facta.junis.ni.ac.rs/me/me201301/me201301-04.pdf
Best regards
dear fellow researcher if you implement the meta-heuristic algorithms of interest on your own, you can easily track the percentage of exploration and exploitation performed by the algorithm during the search. We have done it in our case for solving an optimization problem in Search Based Software Engineering (SBSE). The paper title is 'Fuzzy adaptive teaching learning-based optimization strategy for the problem of generating mixed strength t-way test suites'.
It would be great if you could share the formula you know for the quantitative measure of exploration. Thank you..
The next papers may be useful:
https://ieeexplore.ieee.org/abstract/document/4630938
Article A parameter control method of evolutionary algorithms using ...
Best regards
Hello Seyed,
You raise an important question which has no straightforward answer: how to adequately measure (and drive) the balance between diversification/exploration and intensification/exploitation in a metaheuristic. This question is recurrent in all methods, often a key for success, and connected to the topics of fitness landscape analysis and diversity management. Instead of going into exotic nature-inspired algorithms, as recommended in some previous answers, I recommend to take a step back for a technical analysis of the main components of the methods and problem structure. When focusing on a local search for example, it is meaningful to evaluate the ratio between the total number of solutions of your problem and the number of local minima. This ratio tells you how far your local search is able to "contract/project" the set of solutions into a smaller subset of local minima on which you will conduct the search. Naturally, this ability comes at a cost, in terms of CPU time, and you will have to find a good trade off between local-search complexity and solution quality. Here are a some other literature pointers connected to these topics:
Chapter A Comprehensive Survey on Fitness Landscape Analysis
Article Review: Measuring instance difficulty for combinatorial opti...
Article Metaheuristics in Combinatorial Optimization: Overview and C...
Article Heuristics for Multi-Attribute Vehicle Routing Problems : A ...
Good luck !
-- Thibaut
Hi,
Please read my paper titled A Time-varying transfer function based BPSO, applied soft computing. You will find the answer of your question.
Presumably it should be a function of the dimension of the search space - and it will not be polynomial in that parameter.
You can have a look at the following paper, especially section 3.1, regarding measuring exploration and exploitation.
Article Prey predator hyperheuristic
Fakhrud Din
Thanks Fakhrud Din. For quantitative measure of exploration, you can simply compute the distance of agents from the mean value of the agents.
Thanks Dr. Rafael Rivera-Lopez . The papers helped me a lot.
Thanks Dr. Thibaut Vidal for the useful response. but in some contexts, we can not count the number of local minima..
Thanks Dr. Surafel Luleseged Tilahun for the suggestion. It was so interesting.
- Badges
- Science topic
- Similar topics
- Engineering
- Control Theory
- Control Systems
- Control Systems Engineering