In comparison swarm-based with evolution-based algorithms, the former has some more advantages over the latter. Since evolutionary approaches have relatively more operators than swarm-based, they are more difficult to apply. Different than evolution-based approaches that immediately discard any obtained piece of information related to the old iteration once a new population is generated, swarm-based algorithms normally keep these valuable pieces of information over the subsequent iterations.
Please come in and lower your bucket in this discussion thread.
My two cents is that any heuristic is a heuristic - it will do well in some test cases and will not do well in some. It is better to focus on the optimization problem rather than the method of optimization.
C.P. Ravikumar: I respect your point of view. I have pointed out a similar point of view in my last research paper:
Article Sea Lion Optimization Algorithm for Solving the Maximum Flow Problem
In the paper, I have said the followings:
"Apart from the foregoing mentioned discussion, all metaheuristic optimization approaches are alike on average in terms of their performance. The extensive research studies in this field show that an algorithm may be the topmost choice for some norms of problems, but at the same, it may become to be the inferior selection for other types of problems. On the other hand, since most real-world optimization problems have different needs and requirements that vary from industry to industry, there is no universal algorithm or approach that can be applied to every circumstance, and, therefore, it becomes a challenge to pick up the right algorithm that sufficiently suits these essentials."
Despite the aforementioned discussion, please inform me about your respected point of view about which is the most fertile research area in the field of metaheuristics.
Finally, thank you so much for your rational answer.
In the frame of the Integration of Variables method could be included all then populatio based heuristics, including GQ, PSO and the others. According this method are making evolve solutions options codes helped by any operatori systems in one or more popoulation, depending on the charateristics of each partcular appliation. In my reseqrchgate profile you can find ome old and new algorithms developed.
Best wishes
My one cent goes to the utilisation of mathematical optimisation - every day - and to not even look at metaheuristics. First of all read on the past effort of the problem that you are facing, and if your problem is a new one, and you have some experience in modelling, you probably can create a mathematical model that is suitable. Better still, create more than one mathematical model - as in creating equivalent problems by using other forms of variables or constraint declarations. In large-scale cases, to pick the right form of variables is quite essential for a good result.
I could add that depending on the model structure could be useful the maanipulation of the model by it decomposition, solutions generation by sub-models and the later solutions composition. These operations require using population based solution methods.
I agree with Michael Patriksson and José Arzola-Ruiz . Trying to use an optimization algorithm by all means is a necessary first step: coming up with different formulations, decomposition, and exploiting any special structure that may be present in the problem. However, I disagree with the approach of using metaheuristics for simple problems like the maximum flow, assignment or shortest path problems, or network optimzation problems in general. The example paper that you added in one of the replies uses a "sea lion optimization algorithm" to solve the maximum flow problem. There are numerous fast and efficient algorithms to solve the maximum flow problem, there is no need for metaheuristics. I have written a few papers on the maximum flow problem:
Article A computational study of the capacity scaling algorithm for ...
Article A specialized network simplex algorithm for the constrained ...
Article A faster polynomial algorithm for the constrained maximum flow problem
Article A double scaling algorithm for the constrained maximum flow problem
Article On a capacity scaling algorithm for the constrained maximum ...
I totally agree with Cenk Çalışkan , sometimes using metaheuristics to solve simple problems might not be necessary in some cases.
I also agree with Cenk Çalışkan, but the problem consists in the methodology used for the mathematical formulation and decomposition to assure right models sttructures. secondly it is needed to select right algorithms for solvind each task resulted from the decomposition and the solutions composition algorithm. My work precisely on the solution of these kind of problems
https://www.intechopen.com/books/recent-advances-on-meta-heuristics-and-their-application-to-real-scenarios
- Badges
- Science topic
- Similar topics
- Organizational Psychology
- Training