You cannot really compare two distinct optimization techniques in the general case. Depending on the problem, one technique may perform better than the other. Could you please provide more details about the problem you want to solve ? Please note also that the term "revolutionary" is not appropriate. It should be replaced by "evolutionary".

Agreed. Each technique has specific mechanisms and representational components. In GAs, it's crossover and schemata. In ACO, it's edge-based construction and edge weights. Rather than trying to fit a problem to a specific technique, I would prefer to analyze the problem, identify its key components, and then try to match these components to a suitable metaheuristic.

I agree that there is no one-tool-for-all problems. This link has a helpful guide for identifying candidate algorithms to fit your problem characteristics: http://plato.asu.edu/sub/pns.html

This said, we find some algorithms to be robust or easier to use over a range of problems. SRES (stochastic ranking evolutionary strategies) was found to be one of the easier algorithms to use for global continuous optimization problems in http://scholar.google.com/scholar_url?hl=en&q=http://www.researchgate.net/publication/221526183_Nonuniform_sampling_for_global_optimization_of_kinetic_rate_constants_in_biological_pathways/file/e0b49514d3961a8edb.pdf&sa=X&scisig=AAGBfm26yONJIuvAUuX7SmLYAF_BOC17NQ&oi=scholarr

Your mileage may vary.

Conference Paper Nonuniform Sampling for Global Optimization of Kinetic Rate ...

So, thanks. I,ve received good information. Imagine we have a production function; a function which shows the relation ship between crop yield and energy inputs used in a greenhouse (fertilizers, electricity, gas, etc.). Also we have some constraints. now we want to optimize this function, or imagine we have an objective function which shows the performance of an engine. Now, what do you think? Is it possible to analyse different optimization methods?

I think every algorithm has its special advantage. When using them, we should better design them with respect to the character of given problem. If you designed them carefully, you might find they could achieve similar results

Do you want to solve the problem Or analyze different optimization algorithm on one specific problem?

given your problem mentioned, GA, PSO, CSO, GWO, and even Bee will probably work well if you fit your problem into them.

Ant though is different. It won't easily fit into every problem, If you force it, then its useless even if you get good results.

These are swarm based optimization, algorithms and fuzzy, neural and some newer methods are different.

So what @Zhenzhen Zhang mentioned is really true, if you fit them well, you get similar results though implementation of different algorithms might differ just a little when they fall into one category

What @Stephen Chen mentioned about specifics of problem, is somehow most important.

link provided by @Scott Lett makes it easier to choose a method if one decides to solve the problem in most efficient method and time.

any way, if you want to analyze algorithms by implementing on a specific problem, then after implementation you will find out all mentioned points by yourself.

Why just use these two (ant vs. genetic)? Many optimization problems can be solved using more classic techniques without much overhead than those examples. A great example is using matching theory to attack scheduling problems because those types of problems lend themselves to matching tasks to machines very well and sometimes if you are lucky with your restrictions, you can find a nice means of using computational resources to solve it. So don't limit yourself.

Many optimization problems don't just fit into one technique but are highly related to the mathematical nature they have with their cousin problems (e.g., scheduling and matching problems tend to go hand in hand but need not be the case, I have seen people based on their restrictions get chromatic number problems to line up with scheduling problems). A good scientist notices these things when considering problems in the attempts to find algorithms to solve optimization problems. If you work hard at it, you could prove a new theorem that yields a great algorithm to be described for that problem!

My team has had quite a bit of success on a class of large scale NP-Hard problems such as the constrained Multiple Traveling Salesman Problem. We have utilized fuzzy c-means clustering followed by a Genetic algorithm. Bench marking these solutions have demonstrated good performance as well as computational efficiency. My suggestion is to search the literature to see where both GA and ACO have been used successfully and then classify the type of problems studied in comparison to your needs. I found that it is very important to be "scalable" and that needs to be a consideration if you want to be able to generalize.

Benyamin,

There is famous 'no free luch theorem' for all non-deterministic algoriths. You will always find a problem best solved by a Monte Carlo approach, while all remaining algorithms are supposed to perform better than that. Besides, I think your problem may, and perhaps should, be attacked with a deterministic algorithm. Why not?

ACO is a combinatorial algorithm. so it is good for discrete problems. GA is good for continuous and discrete. A new version of ant colony was developed for continuous space but it is somehow complicated and hard to use. But I think GA or other heuristic methods such as harmony search would be a better choice for continuous domain problems.

Your question is very general. ACO and GA can work in both discrete and continuous domains. But as a comparative expriment you can have a glance at this paper

"Ant Colony System: A Cooperative Learning Approach to the Traveling Salesman Problem" from Marco Dorigo

In this paper ACO shows a much better performance than GA in solving TSP.

check it out:

http://www.upo.es/eps/troncoso/Citas/ICAE10/citaICAE-2.pdf

I think GA perform better then Ant colony optimization

Relation between Genetic algorithm and ant colony

http://www.google.co.in/url?sa=t&rct=j&q=&esrc=s&source=web&cd=2&ved=0CDgQFjAB&url=http%3A%2F%2Fwww.cnc.una.py%2Fcms%2Finvest%2Fdownload.php%3Fid%3D199695%2C86%2C1&ei=aK2RUv_uCIiNrQeepoHIAg&usg=AFQjCNGtfs1yYJgeWXBal15non_GvDz8Ng&sig2=0sxQAVww_cjETwfcEQSsRg&bvm=bv.56988011,d.bmk

I think the suitable method depend on the type of the optimization problem itself. However, in the last decade there are many algorithms that work on using a hybrid algorithms for example Ant-Genetic algorithm and ant-bee algorithm...etc

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)

Most of the time GA is better than ACO, but again it comes on the problem nature.