2
$\begingroup$

A metaheuristic is defined as

a metaheuristic is a higher-level procedure or heuristic designed to find, generate, or select a heuristic (partial search algorithm) that may provide a sufficiently good solution to an optimization problem

So my question is if a neural net could be consider in that definition

$\endgroup$
1
$\begingroup$

The notion of a "metaheuristic" is not something with a precise formal definition. Personally, a neural network isn't something I'd consider a metaheuristic; it doesn't seem quite like what we have in mind when we talk about metaheuristics. But you can always use language however you like, as long as it is clear to others and doesn't cause confusion.

$\endgroup$
1
$\begingroup$

To add to the other answer(s), you can certainly use neural networks (or machine learning in general) in conjunction with metaheuristics. Indeed, there are several such approaches in the literature, for example, in trying to predict "good" initial solutions for some search method.

$\endgroup$
0
$\begingroup$

When you train NN all things you do is considered methaheuristic, if you use search among NNs to pick best it is also metaheuristic.

In this context, simple usage of NN is not considered as such, but people rarely split hair over definition. What is left is usage, when NN is used as screening of solutions or other methods (cascade of nonhomogenous nets) then it is methaheuristic, otherwise it is simply search/classification tool.

$\endgroup$

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.