It is very positive that people are trying to learn from the nature and try to invent same new algorithms. As a result occasionally we get a valuable new algorithm. But sometimes it can go wrong and people manage to publish "bad" algorithms.

An interesting talk about "new algorithms" was given by Günter Rudolph on ANTS 2014 conference (http://iridia.ulb.ac.be/ants2014/invited_talks.php)

He have analyzed some relatively new algorithms and presented to us (participants) results for same algorithms that are actually only reinterpretation of an old and well known algorithms. Here is the list of algorithms that he consider as old algorithms in disguise:

* Firefly algorithm (Xin-She Yang)

* Bat algorithm (Xin-She Yang)

* Cuckoo search (Xin-she Yang and Suash Deb)

* Shuffled frog-leaping algorithm

* Harmony search

If I remembered it correctly, his analyses showed that Firefly algorithm is a form of genetic algorithm, and Bat algorithm is a kind of PSO with (1+1) step.

The analyses were done by transforming original equations in other (equivalent) forms from which it is more easy to see that the new algorithm is actually not new but some well-known algorithm.

Sometimes the parameter from a well-known algorithm is disguised is more complicated expression. For example to paraphrase: the original algorithm uses parameter A, but the new algorithm is using (1+exp(-B)) instead, where B is a new parameter. It is easy to see that complicated expression (1+exp(-B)) is only inefficiently simulating the old parameter.

He also notes that these “new” algorithms are not following natural phenomena that allegedly inspired the algorithm.

So if you manage to invent same new algorithm, for example inspired by the baby that is searching for a breast it would be important that it is using some new concept that is not seen in other nature inspired algorithm. Of course you can combine concepts from two or more well-known algorithms (hybridization), but as I understand you would like to invent some new nature inspired algorithm.

Some argue that you should find an optimization problem for which this new algorithm is better than the state of the art method before you publish. In this I strongly disagree! I think that if the algorithm have some new technique that is promising, but yet cannot reach the state-of-the-art algorithms it is still worth exploring and publishing. An argument if favor of this is that early variants of Ant Colony Optimization algorithms (AS and EAS) could not compete with the state-of-the-art algorithms, but they were published. As a result of further researches, they were improved and tuned so their successors are very successful and important algorithms.

Don’t be discouraged by these negative examples, because we all know about many positive examples!

Good luck.

Dear friend

Yes there are some recent developments in this direction.

You will see better algorithms and some improvements over them.

Please some of the recent literature.

In fact, researchers are developing algorithms most suitable for their problems at hand.

you may develop a algorithm according to your thought process, just see if it works fine over the existing algorithm by testing over some benchmark data sets. I wish you all the best. I developed some hybrids algorithms that outperform the existing ones. Just you may see some of my publications.

I hope this helps.

Best regards

Dr.Indrajit Mandal

Hi,

the list would be rather long. There are a lot of variations and hybrid approaches of the existing algorithms but one obvious algorithm you missed is the standard genetic algorithm.

As for the "newborn infant movement" one would first need to understand how the newborn moves. Find publications explaining the nature of the process and build upon that that idea.

See this paper:

http://onlinelibrary.wiley.com/doi/10.1111/itor.12001/abstract

In your list some algorithms are missing..

> harmony search

> frog jump

> jumping gene algorithem

> spider monkey

....

all the best for your New Research.

Hi Arie,

the vision is interesting, try to understand the various algorithms available till date and  then develop the architecture for the new one. you need to implement the physics.

All the best!!!

Regards,

Rajesh.

Thank you very much all. Especially for Adam, your recommended paper is very interesting point of view. Thank you

Hi Arie,

I also advise the following paper:

Boussaïd, I., Lepagnot, J., & Siarry, P. (2013). A survey on optimization metaheuristics. Information Sciences, 237, 82-117.

One important thing in developing a metaheuristic is to understand its mechanism. All metaheuristics have two tools in their structure: 1) intensification and 2) diversification. Intensification tool tries to guide the algorithm to the possible fittest solution by highlighting best solutions in terms of objective function. Contrarily, diversification tool tries to create purely random changes in solutions in hope of more exploring solution space and escaping local optimums. 

For example, GA, using Roulette Wheel and crossover operator, gives more chance to better solutions to be selected and produce the next generation's children (which are similar to their parents) (See genetic principle: better individuals have more chance to survive and reproduce). While it uses mutation operator to make random changes in the solutions structure to search other regions of the solution space. 

Or consider SA, at the beginning of the algorithm, chance of selecting worse neighbors is high (using Boltzmann function) and it makes that the worse solutions have a chance to be selected (this mechanism acts as a diversification tool). As the algorithm goes on this probability reduces and the SA turn out to be a simple local search or hill-climbing algorithm (only when a neighbor is selected that it is better than the current solution). Finally the algorithm converges to a local optimum (maybe global optimum) and it finishes. 

If you look other metaheuristics, all of them have embedded intensification and diversification tools in themselves. 

I hope me information will help you yo develop your own metaheuristics based on these two facts.

best wishes.

Not very natural but metaheuristical as well: Simulated annealing

Flower Pollination Algorithms

Cat swarm

Bacterial foraging algorithm

and many more you can find here

https://www.elsevier.com/books/nature-inspired-optimization-algorithms/yang/978-0-12-416743-8

From nature-inspired hyper-heuristics, there is an autoconstructisve mechanism;  Push and SM-CGP are some examples of them. In simple words, the solvers are responsible to reproduce during the evolution process. It is considered an online learning mechanism, that is should help the search for an algorithm  to converge towards an optimum solver quicker. 

Hi                                                                                                                                                        See this paper:"A Brief Review of Nature-Inspired Algorithms for

Optimization"

It is very positive that people are trying to learn from the nature and try to invent same new algorithms. As a result occasionally we get a valuable new algorithm. But sometimes it can go wrong and people manage to publish "bad" algorithms.

An interesting talk about "new algorithms" was given by Günter Rudolph on ANTS 2014 conference (http://iridia.ulb.ac.be/ants2014/invited_talks.php)

He have analyzed some relatively new algorithms and presented to us (participants) results for same algorithms that are actually only reinterpretation of an old and well known algorithms. Here is the list of algorithms that he consider as old algorithms in disguise:

* Firefly algorithm (Xin-She Yang)

* Bat algorithm (Xin-She Yang)

* Cuckoo search (Xin-she Yang and Suash Deb)

* Shuffled frog-leaping algorithm

* Harmony search

If I remembered it correctly, his analyses showed that Firefly algorithm is a form of genetic algorithm, and Bat algorithm is a kind of PSO with (1+1) step.

The analyses were done by transforming original equations in other (equivalent) forms from which it is more easy to see that the new algorithm is actually not new but some well-known algorithm.

Sometimes the parameter from a well-known algorithm is disguised is more complicated expression. For example to paraphrase: the original algorithm uses parameter A, but the new algorithm is using (1+exp(-B)) instead, where B is a new parameter. It is easy to see that complicated expression (1+exp(-B)) is only inefficiently simulating the old parameter.

He also notes that these “new” algorithms are not following natural phenomena that allegedly inspired the algorithm.

So if you manage to invent same new algorithm, for example inspired by the baby that is searching for a breast it would be important that it is using some new concept that is not seen in other nature inspired algorithm. Of course you can combine concepts from two or more well-known algorithms (hybridization), but as I understand you would like to invent some new nature inspired algorithm.

Some argue that you should find an optimization problem for which this new algorithm is better than the state of the art method before you publish. In this I strongly disagree! I think that if the algorithm have some new technique that is promising, but yet cannot reach the state-of-the-art algorithms it is still worth exploring and publishing. An argument if favor of this is that early variants of Ant Colony Optimization algorithms (AS and EAS) could not compete with the state-of-the-art algorithms, but they were published. As a result of further researches, they were improved and tuned so their successors are very successful and important algorithms.

Don’t be discouraged by these negative examples, because we all know about many positive examples!

Good luck.

The procedure of most evolutionary algorithms is as the same. I mean they have exploration and exploitation and they all share many common points but, the operations make them different with each other. If we compare the

information sharing of GA, PSO and ACO, we will see that information

sharing of them are significantly different. You may consider the following optimization algorithm as well, Forest Optimization Algorithm (FOA)  which is inspired by forests:

http://www.sciencedirect.com/science/article/pii/S0957417414002899

good luck

Hi Arie, 

Following paper contains nature-inspired in particular theories of physics inspired algorithms with  there variants and various applications. 

https://www.researchgate.net/publication/237101455_Physics-Inspired_Optimization_Algorithms_A_Survey

Hope this would help you.

Article Physics-Inspired Optimization Algorithms: A Survey

Hello 

I feel every researcher prove that his or her implemented algorithm is best because the nature inspired algorithms work with random theory. There is no suitable algorithm which can be proved as global best algorithm because every inventor of an algorithm propose that his algorithm outperform the existing one,  maybe they do manipulation in their results to prove his/her work best. May be I can be wrong but this was all my personal experience that I feel.

Sorry researchers if you get discourage from above negative points. And if any one agree with me then please share his or her views here.

It is an interesting comment from Shonak. This type of impression could be  reached from the literatures. I suppose if somebody has spent a long time, writing an algorithm, this person should be proud of his/her work. Manipulating results is not a good research practice, really. Perhaps we all have this desire to show our results are good and we may design our experiments to demonstrate them. At the end of the day, results should be reproducible.

I focus my research on letting the computer to optimise the algorithms and generates new algorithms from scratch. With this method, the algorithms are not limited to our human-mind, which could be sometimes restrictive. In my case, I let a nature-inspired system to search for the algorithm. 

Interesting conversation.

Patricia  

Right Patricia, can you tell which one algorithm is best????

Forget the nature, the birds and the bees; what do they know about optimization? Well, probably they do, evolution does, at least. What most known heuristics do is trying to adjust a natural process to the solution space of the problem at hand. Would not be more efficient to adjust the optimization process to the solution space? Here is an attempt:

http://web.unbc.ca/~bluskovi/pdo.pdf

Journal of Combinatorial Optimization, 2015, DOI 10.1007/s10878-014-9826-x.

The final publication is available at link.springer.com; more precisely, here:

http://link.springer.com/article/10.1007/s10878-014-9826-x

Supplementary material (illustration on how PDO works in one possible instance; finding cyclic designs):

http://static-content.springer.com/esm/art%3A10.1007%2Fs10878-014-9826-x/MediaObjects/10878_2014_9826_MOESM1_ESM.pdf

Long live the problem itself!

Hyperheuristics optimises algorithms to solves problems. A stochastic method generates a new meta-heuristics. It is a step forward from optimising the parameters.

I would like to recommend you the following article: I may be helpful to you..

Mahamad Nabab Alam, Biswarup Das, Vinay Pant, A comparative study of metaheuristic optimization approaches for directional overcurrent relays coordination, Electric Power Systems Research, Volume 128, November 2015, Pages 39-52, ISSN 0378-7796, http://dx.doi.org/10.1016/j.epsr.2015.06.018.

(http://www.sciencedirect.com/science/article/pii/S0378779615001947)

https://www.researchgate.net/publication/296636431_Codes_in_MATLAB_for_Particle_Swarm_Optimization?ev=prf_pub

Code Codes in MATLAB for Particle Swarm Optimization

https://www.researchgate.net/publication/297245624_Particle_Swarm_Optimization_Algorithm_and_its_Codes_in_MATLAB

Research Particle Swarm Optimization: Algorithm and its Codes in MATLAB

https://www.researchgate.net/publication/280244004_A_comparative_study_of_metaheuristic_optimization_approaches_for_directional_overcurrent_relays_coordination

Article A comparative study of metaheuristic optimization approaches...