We are facing a true tsunami of “novel” metaheuristic. Are all of them useful? I referred you to Sorensen - Intl. Trans. in Op. Res. (2013) 1-16, DOI: 10.1111/itor.12001.
I agree with you, I think that the most novel metaheuristics are similar, the most of them can be viewed as variants of GA, PSO, or ACO.
The common feature of "novel metaheuristics" as you call them, is lack of determinism. I wouldn't say that this fact alone moves them away from scientific rigor, but indeed, there may be a problem with estimating the accuracy of solutions obtained on this way. We like Genetic Algorithms since they can sometimes "invent" quite unexpected qualitative results, better than those we already know. Is their "creativity" scientific or rigorous? I don't know, but it would be stupid not to use them.
I want to offer my view on this important question.
Before I want to write a correlation LB(D) A_j(D2, T1); (c) A_i(D1, T2) > A_j(D1, T2); (d) A_i(D2, T2) < A_j(D2, T2); where D1 and D2 are two different data sets, T1 and T2 are two different time limits. Thus, for arbitrary data set D and time limit T we can get different "best" of algorithms A_i: A_i(D, T) = min { A_1(D, T), A_2(D,T), ... A_n(D, T) }. Let given the following problem: to find any solution A_i(D, T), 1
The interplay among Fedulov and other authors, namely the Mihai Banciu, is very interesting. However, it seems me that Fedulov does not answer to my question. I cite Sorensen (2013 - DOI: 10.1111/itor.12001) below to emphasize/complement my start question:
"... surge of “novel” methods seems to have very little on offer besides a new way of selling existing ideas. The development of ever more “novel” metaheuristic methods has another disadvantage, in that it distracts attention away from truly innovative ideas in the field of metaheuristics."
Dear Nuno Costa,
In my view, ideas of finding quality lower bounds of objectives functions can be considered as "truly innovative ideas in the field of metaheuristics". I think, as before, that finding quality lower bounds is many times (maybe million times) more important thing than finding quality metaheuristic solutions. My explanation is the following.
Finding quality metaheuristic solutions is more craft than science because of unpredictability of output results. But finding quality lower bounds of objective functions is only science, since gives strict proof that there doesn't exist any solution for x < Lower Bound, where x is any value of objective function.
there are people that thinks so. The reason is that apear every week a new algorithm which seeasm to be a copy from other only with different names