Combining different (meta-)heuristics results in something that is known as hybrid (meta-)heuristic. There is a lot of literature on how to properly hybridize two (or more) classical (meta-)heuristics, and also an annual workshop on this topic, see http://iwi.econ.uni-hamburg.de/hm14/ for last year's workshop.

Dear Rahman,

Various researches have used a combination of meta-heuristics algorithms. For instance, Miranda presents the DE-EA-PSO in his paper (http://ieeexplore.ieee.org/xpl/articleDetails.jsp?arnumber=6855877&pageNumber%3D33561%26rowsPerPage%3D100), improving the results. Another paper (by Lee et al.) shows GA-ACO algorithm has superior performance when compared to other (http://www.sciencedirect.com/science/article/pii/S1568494606000822).

It is important to understand the problem (mathematical model) very well to adjust the better approach. So, you can enjoy the advantage that each chosen meta-heuristics offers.

Combining different (meta-)heuristics results in something that is known as hybrid (meta-)heuristic. There is a lot of literature on how to properly hybridize two (or more) classical (meta-)heuristics, and also an annual workshop on this topic, see http://iwi.econ.uni-hamburg.de/hm14/ for last year's workshop.

Hi Rahman,

it is not only possible to combine different metaheuristics, it is also recommended.  A great deal of research has recently been dedicated to the so called hybrid metaheuristics.

For further details, I would advice you to read the article that I attached to this message as well as the excellent book edited by Christian Blum et al. : Hybrid metaheuristics.

Hybridizing different metaheuristics, when done properly, allows you to obtain an enhanced approach that combines the qualities and advantages of each of the combined methods but also reduces the impact of their respective drawbacks.

I hope this helped,

Best regards,

Imène.

yes, it is called hybrid meta-heuristics, however, you have to be careful as which algorithm to merge with which one as the complexity will get harder if you try to merge two complex algorithms, or part of the algorithms. The main aim in doing this is to balance between the exploration versus the exploitation behavior of these algorithms. Sometimes, it is better if you know how to do it and why you are doing it.

According to the No Free Lunch Theorems for Optimization, there is no single algorithm that is suitable for all kinds of optimization problem. An algorithm may perform well in the initial stage of the problem solving process but later its performance will be bad. So, nowadays, the direction is to use more than one algorithm to solve a problem, and it came in many forms for example; hybrid EA or hybrid meta-heuristics, Multi-method, Multi-operator,... etc 

But my experience says.. although it is possible to develop a hybrid meta heuristic method..but at the cost of speed of convergence..well my application is limited to some power system optimization cases only hence cannot be generalized

Hyper heuristic and genetic improvement  can generate hybrid algorithms that are specialised to solve instances of a problem.

 Efafe

As previously mentioned with regards to No Free Lunch, each metaheuristic has advantages and disadvantages.  Subsequently, ANY hybrid can be expected to have some advantages on some problems.  Our line of research has been to focus on analyzing the search characteristics of various metaheuristics to determine some guidelines as to how and when a hybrid should be implemented. 

In the attached paper, we analyze several metaheuristics and several hybrids.  We then show how proper selection of the hybrid components can lead to better overall performance.

Conference Paper Multi-swarm hybrid for multi-modal optimization

You may also check our work where we integrate two metaheuristic techniques, GA and SA (attached).

Conference Paper An integral approach for Geno-Simulated Annealing

It it possible. There is a special area related to this topic called hyper heuristics. This is a survey where you can find recent proposals and their applications.

http://www.palgrave-journals.com/jors/journal/v64/n12/full/jors201371a.html

Of course. We called them cooperative strategies for optimization. Can also be named as portfolio algorithms. Check my publications at decsai.ugr.es/~dpelta

Sure. It is true that we have some well-known metaheuristic approaches in the literature, but it does not mean that you cannot change their structures. In almost every cases you should engineer your own metaheuristic approach to solve a certain problem. To do this you can also combine different algorithms or characteristics of those.

Of course. We called them Hybrid algorithm and we can combine a global algorithm (GA, PSO, ACO..) and a local algorithm (SA, Tabou...).

Please see to this chapter:

M. Fakhfakh, A. Sallem, M. Boughariou, S. Bennour, E. Bradai, E. Gaddour and M. Loulou, "Analogue Circuit Optimization Through a Hybrid Approach", Studies in Computational Intelligence: Hybrid Evolutionary Algorithms, Springer 2011.

Regards.

Amin

Combining various features of more than one metaheuristic that perform together and complement each other may result in more robust combinatorial optimization tools. You can see the works of  Blum and Roli, 2003 and Raidl, 2006. Hybridization, when properly applied, may further enhance the effectiveness of the solution space search, and may overcome any inherent limitations of single metaheuristic algorithms. Therefore, new opportunities emerge, which may lead to more powerful and flexible solution methods for combinatorial optimization problems.

Regards

Sure, this is quite possible and there are lots of studies on hybrid Meta-heuristics. Simply search google scholar on  'hybrid Meta-heuristics' for a comprehensive literature on this subject

The above answers are quite complete. I would recommend you to use a form a genetic programming, especially a graph based ones like CGP. It does not bloat and it easier to use.

Yes, off course....

Dear researcher and I hope all doing well,

Can anyone explain how can hybrid two algorithms to enhance and perform the system eg, if you have a problem and I have a code of Matlab for the optimization with GA and other code of Matlab already with PSO, how can do that?

thanks for the answer in advance.

Amin Sallem

Fatima Benbouzid-Si Tayeb

Patricia Ryser-Welch

Julius Odili

Humyun Fuad Rahman

Mahamad Nabab Alam

Yes, you can use the best features of any two algorithms. In this way, you may able to improve performance.

Mahdi Abdeddaim

Certainly, you can combine heuristics by stopping one and initiating another at some point. One may be better at diversifying the search and another might be better at intensifying the search. The "art" is to figure out when to change search heuristics, and for how long one heuristic will operate - then whether to switch back and so on.

Yes, you can combine key steps of one algorithm with key steps of the other algorithm, the resulting algorithm is the namely hybrid algorithm, which has outstanding performance as compared to their original versions. For examples, you can apply crossover and mutation of the genetic algorithm on any other algorithm like PSO, ACO, BFOA, etc

Yes, You can find the performance of meta-heuristic algorithm hybridization such as combining PSO/PPSO with GSA, >>>PPSOGSA.

10.1109/ACCESS.2019.2909561

I recommend reading the following books

" Combining Machine Learning and Metaheuristics Algorithms for Classification Method PROAFTN"

Chapter Combining Machine Learning and Metaheuristics Algorithms for...

Definitely. We are doing this since many years ago. We called it "cooperative strategies". And in this paper from 2013

Article A centralised cooperative strategy for continuous optimisati...

we showed why they obtained better results. Essentially, they can reach regions of the search space that cannot be reached by the individual algorithms.

Regards

David

Yes, it is certainly possible. We can think of an ensemble meta-heuristic technique that combines the results of various independent meta-heuristics as done in Machine Learning. It can provide more accurate solutions to an optimization problem than a single method would.

Remember the No-Free lunch theorem.... If once. combination works for a set of problem, it may not for another. So, I would refine your question and states the types of problems you are. attempting to solve.

In my view, you can combine multiple-algorithm. but, you should know that if you couldn't combine well, may be the results traps on locals. For example you can use 50% percent of your pupation in GA and 50% in PSO and update it at the end of each iteration. I don't do this work, but It may be interesting.