2
$\begingroup$

In a research paper i have written i am solving a problem through the use of an ANN, which learns to model a technical process through training data.

I compare that to metaheuristic approaches, which can also find relevant outputs from inputs by searching, e.g., with a genetic algorithm and an appropriate cost function.

A reviewer gave me the feedback that ANN are also a kind of metaheuristics, so i should not compare ANN vs. metaheuristics as different kinds of approaches in the paper. Is this correct? I have looked up the definition of metaheuristic and while i think that i could use a metaheuristic optimizer instead of a gradient based one, an ANN itself is not a metaheuristic.

Improve this question
$\endgroup$
1
  • $\begingroup$ If you think of a heuristic as a solution to a particular problem (can also be seen as a cost function), then a metaheuristic is a heuristic that is independent of application (very broad scope). ANNs are pretty broad in their applications and they propose a generalized methodology of data-driven modeling, therefore I can see the analogy between them and metaheuristics. $\endgroup$
    – pcko1
    Jun 21, 2018 at 8:05

1 Answer 1

Reset to default
2
$\begingroup$

I suppose one could possibly argue it either way, but at the end of the day, it doesn't really matter whether we classify an ANN as a metaheuristic. It sounds like you have one reader who found it off-putting or confusing to have you compare ANN vs metaheuristic because they would have characterized an ANN as a metaheuristic. So, I suggest asking yourself three questions: (1) How many other readers will have a similar reaction? (2) What changes can I make to the paper that would avoid that reaction? (3) Is there any downside to those changes? Are they worth it?

In this case, it seems easy enough to rephrase your description of the comparison, so you don't trigger this reaction. You could be more specific about what comparison you are doing: you compare an ANN to a genetic algorithm. That might avoid the problem, with no downside. Being specific is often a good thing.

Improve this answer
$\endgroup$
2
  • $\begingroup$ Thanks for the answer. I know how to appropriately react to the comment and change the content, but i was still wondering if the reviewer had a good point. $\endgroup$
    – JanM
    Jun 22, 2018 at 6:47
  • $\begingroup$ @JanM, OK, makes sense. I wouldn't have considered a neural network as a metaheuristic, personally, but I am skeptical whether this is something where there is a clear objective notion of correct and incorrect. To me, the concept of "metaheuristic" lacks a precise definition and seems open to interpretation, so trying to debate exactly what it means seems like a waste of time to me. $\endgroup$
    – D.W.
    Jun 22, 2018 at 7:18

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.