How to design an equation's (algorithm) by using nature, what is the principle behind these algorithm (PSO, firefly, cuckoo search, BAT algorithms, etc..), how they are creating ? as far now i have studied all these algorithms but i don't have any clear idea what's the logic behind it. i'm trying a lot but i can't, i need help...
Firstly, I would like to suggest that you start with a natural process that is doing optimization. Foraging ants, fine. Fireworks? Leaping frogs? Musical harmonies? These are better suited as prime examples of what not to do (see Sorensen's paper attached).
Secondly, why not start with a specific optimization situation and then look for a natural analog? Deciding when to explore a new attraction basin or stay to exploit the current one? Simulate the process of annealing!
Good luck! (And there's a reason why most of the metaheursitics developed in the last 20 years are on the fringes of the research community...)
Article Metaheuristics -- the metaphor exposed
They are metaheuristic algorithms:
In each iteration of these algorithms:
- The best solution is found
- Other solutions move toward the best solution by specified rules
- The position of few solutions may be changed in stochastic mode (Like mutation operation in GA) [Some algorithm like PSO do not have this operation]
The general idea behind them is employing swarm intelligence. Firstly, a swarm-based activity to achieve an objective must be detected. Then, a mapping strategy might be required to computerize that natural activity and create an optimization framework.
In simple words, all of you need is a beautiful mind :)
Hi, Good question. These links can help you more.
http://dl.acm.org/citation.cfm?doid=1118178.1118215
http://msb.embopress.org/content/7/1/546
http://dl.acm.org/citation.cfm?id=1506427
Dear Arul,
Metaheuristic Algorithm whether nature inspired or bio inspired has two properties intensification and diversification. Intensification is the process when you find a good region of the search space, you examine it carefully to see if it contains a good local optima, on the other side diversification introduces new solution in the search space. Sometimes they are called "Exploitation" and " Exploration ", respectively. Designing a new metaheuristic algorithm we need to follow the points.
- Set of assumptions (rules) which simplify the natural/biological phenomena.
- Mapping of natural /biological phenomena with the intensification and diversification through some mathematical equations.
- Set of control parameters which will control the intensification and diversification process.
- Selection of standard objective function (Sphere, Booth, Schwefel, Rosenbrock, Six-Hump Camel etc) on which you’ll apply the algorithm.
- Compare your algorithm with other existing algorithms. Here need fine tuning of parameters.
I hope these points can help you.
Many processes of the nature are efficient. So, the goal is to get an adequate model of the nature process but this is not a easy task.
Basically, Nature inspired optimization algorithms inspired from either Biological, Physical or Chemical phenomenon.
The core idea in any nature inspired optimization algorithms is the local search and global search. Means for the development of any new nature inspired optimization algorithm we need to devise two operators one is exploitation or intensification (local search ) and other is exploration or diversification (global search) inspired from the natural phenomena. These two operators are useful to find the near optimal solution within a given time.
Firstly, I would like to suggest that you start with a natural process that is doing optimization. Foraging ants, fine. Fireworks? Leaping frogs? Musical harmonies? These are better suited as prime examples of what not to do (see Sorensen's paper attached).
Secondly, why not start with a specific optimization situation and then look for a natural analog? Deciding when to explore a new attraction basin or stay to exploit the current one? Simulate the process of annealing!
Good luck! (And there's a reason why most of the metaheursitics developed in the last 20 years are on the fringes of the research community...)
Article Metaheuristics -- the metaphor exposed
Stephen Chen- wow! This Article took's my research into next level , Thank you For sharing, i think every one should read this article it's really interesting (even i have Conway-ed my personal acknowledged to this author).
Thank you,
With Regards,
-Arul
- Badges
- Science topic