Talk:Metaheuristic

From Wikipedia, the free encyclopedia

WikiProject Systems
Systems science Portal
This article is within the scope of WikiProject Systems, which collaborates on articles related to Systems science.
Systems rating: B Class Mid importance  Field: Operations research
Please update this rating as the article progresses, or if the rating is inaccurate. Please also add comments to suggest improvements to the article.

Contents

[edit] Heuristic search in AI

There are lots of works on heuristic search in Artificial Intelligence which are not mentioned in the article and definitely belong in it, including A* and its extensions. —Preceding unsigned comment added by 203.143.165.77 (talk) 06:39, 20 May 2008 (UTC)

[edit] Merge

Diomidis Spinellis believes this page would be better off at metaheuristic. I agree. However, the approach Diomidis Spinellis took, by cut and paste, is not the right one, since that destroys the page history.

I would like to ask an admin to delete the redirect now at metaheuristic, then move this page there. The links to this page were already taken care for. Thank you. Oleg Alexandrov 17:37, 27 Apr 2005 (UTC)

  • Agree. Corrects a glaring usage error. ---Isaac R 17:55, 27 Apr 2005 (UTC)
Done. Lachatdelarue (talk) 22:11, 28 Apr 2005 (UTC)
Dear Lachatdelarue, thanks! Oleg Alexandrov 23:03, 28 Apr 2005 (UTC)

[edit] Citing sources

Hi All. It is unclear whether some of the text in the article is the writers’ thoughts, or generally accepted fact. As an encyclopedia, citation is necessary to back up arguments. --Daleh 12:29, 27 September 2006 (UTC)

The opinions in the text are clearly not generally accepted. In fact, they seem to come from someone who hates metaheuristics. Evolutionary computation researchers such as myself take serious issue with statement such as that no particular class of metaheuristics does better than any other for particular problem classes. Also, many of us are not very convinced of the wide-ranging implications of the no free lunch theorem. Daleh

[edit] Clear bias and misleading discussion

This article contains a clear bias against metaheuristic based approaches. I find the language used is especially emotive and far from an objective description or assessment of the concept.

I also find the final paragraph of the "General criticisms" section to be highly misleading. Whilst the paragraph is factually correct in its statements regarding continuous non-linear optimisation, it implies these methods should be easily applicable in metaheuristics. Given that metaheuristics are primarily used for combinatorial optimsation and constraint satisfaction problems where the solution space is discrete I find this to be a highly dubious and un-supported inference.

Axiomaticus 03:05, 8 May 2007 (UTC)

[edit] More criticisim

I agree with Axiomaticus, this article is presented in a very biased manner that would seem to me to be founded in the authors limited perspective on metaheuristic methods. In addition it is not, in my opinion, even factually correct. For instance evaluating the hessian at a particular point in space to obtain a good search direction works very well on some problems, but for others where the search space is non-smooth or discontinuous or for some other reason cannot be approximated using for instance, a taylor series expansion about a point or using some other perturbation methods, the hessian will be difficult to evaluate and may be a very misleading.

With regard to the comments about spaces deceptive to metaheuristic algorithms, and to the no-free-lunch theorem. The no free lunch theorem applies to all algorithms and all problems, the it is as easy to find a function to fool a second order method, or to cause slow convergence as it is to for any algorithm.

--87.192.57.3 12:14, 22 May 2007 (UTC)Conor