Talk:Evolutionary algorithms

I'm not fully satisfied with this short introduction to Evolutionary Algorithms (EAs).

Given the fact that this should be a general introduction, parts of the text are too much focused on special topics while other aspects are only mentioned marginally. The text contains a lot of references, however, a considerable amount is of very special nature or is somewhat outdated and should not appear in an introductory article. Also, according to the "Instructions for authors" the number of references should be kept to a minimum, i.e., "citing only major books and review articles." Therefore, the author should look at each citation listed whether it meets this requirement.

I will discuss the sections consecutively.

According to the general Scholarpedia recommendations, the start of the article should be an "one-paragraph dictionary-like definition." Therefore, the second part beginning with "Darwinian evolution ..." should be removed or partially incorporated in the first paragraph or moved to another place in the article. Furthermore, this should be the location in the article where the main branches of EAs *must* be mentioned, i.e., evolutionary programming, evolution strategies, genetic algorithms, and genetic programming. Additionally, these main EA branches must be linked to the corresponding Scholarpedia articles. Furthermore, each of these branches should have one reference to a text book or overview article. For example (using my bibtex database, you might find other references better suited):

@BOOK{Fog95,

   author  = {Fogel, D.B.},
   title   = {Evolutionary Computation},
   year    = {1995},
   publisher = {IEEE Press},
   address = {New York}

}

@ARTICLE{BS02,

       AUTHOR          = {Beyer, H.-G. and Schwefel, H.-P.},
       TITLE           = {Evolution Strategies: 
                           A Comprehensive Introduction},
       PAGES           = {3--52},
       JOURNAL         = {Natural Computing},
       VOLUME          = {1},
       NUMBER          = {1},
       YEAR            = {2002}

}

@BOOK{Gol02,

       AUTHOR          = {Goldberg, D.E.},
       TITLE           = {The Design of Innovation: Lessons from and for 
                           Competent Genetic Algorithms},
       PUBLISHER       = {Kluwer Academic},
       ADDRESS         = {Boston},
       YEAR            = {2002}

} @BOOK{Koz92,

       AUTHOR          = {Koza, J.R.},
       TITLE           = {Genetic Programming: On the Programming of 
                           Computers by Means of Natural Selection},
       PUBLISHER       = {MIT Press},
       ADDRESS         = {Cambridge, MA},
       YEAR            = {1992}

}

The section "Basic Approach" should also mention hybrid algorithms sometimes referred to as memetic algorithms, estimation of distribution algorithms, and the concepts of distributed population models. The final sentence "Many refinements ..." is a somewhat chaotic collection of special and general concepts containing references to very special topics. This part should be rewritten in such a manner that it explains the "refinements" more detailed such that the reader can grasp the basic ideas. Also the literature references should point to *general* (i.e., introductory) papers and not to very special topics. For example, as to adaptation techniques, the following reference could be used:

@article{EHM99,

       AUTHOR          = {Eiben, A.E. and Hinterding, R. and Michalewicz, Z.},
       TITLE           = {Parameter Control in Evolutionary Algorithms},
       JOURNAL         = {IEEE Transactions on Evolutionary Computation},
       VOLUME          = {3},
       NUMBER          = {2},
       YEAR            = {1999},
       PAGES           = {124--141}

}

or alternatively

@BOOK{LLM07,

       EDITOR          = {Lobo, F.G. and Lima, C.F. and Michalewicz, Z.},
       BOOKTITLE       = {Parameter Setting in Evolutionary Algorithms},
       PUBLISHER       = {Springer},
       ADDRESS         = {Berlin},
       YEAR            = {2007}

}

I strongly recommend to reduce the number of references to a minimum of basic overview papers. The author should check his specific choices in detail.

The section "Origins" is the one that I really like. This is the only place where I would recommend to keep all references.

The section "Basic Mathematical Properties" is the largest one. This is in contrast to the author's statement "These theoretical findings are, however, not very practical". Why wasting so much space for things not very useful? Again, the citation presented are very specific (by the way, "B"ack and Hoffmeister, 2004" has wrong year, it must be 1994). Meanwhile there are books on that topic that might be cited instead:

@BOOK{Bey01,

       AUTHOR          = {Beyer, H.-G.},
       TITLE           = {The Theory of Evolution Strategies},
       PUBLISHER       = {Springer},
       SERIES		= {Natural Computing Series},
       ADDRESS         = {Heidelberg},
       YEAR            = {2001}

}

@book{Spe00,

       AUTHOR      	= {Spears, W.M.},
       TITLE      	= {Evolutionary Algorithms: The Role of Mutation 
                           and Recombination},
       PUBLISHER       = {Springer-Verlag},
       ADDRESS         = {Heidelberg},
       YEAR            = {2000}

}

@book{Vos99,

       AUTHOR          = {Vose, M.D.},
       TITLE           = {The Simple Genetic Algorithm: Foundations and 
                          Theory},
       PUBLISHER       = {MIT Press},
       ADDRESS         = {Cambridge, MA},
       YEAR            = {1999}

}

The overview article could be used alternatively

@ARTICLE{BSW02,

       AUTHOR          = {Beyer, H.-G. and Schwefel, H.-P. and Wegener, I.},
       TITLE           = {How to Analyse Evolutionary Algorithms},
       PAGES           = {101--130},
       JOURNAL         = {Theoretical Computer Science},
       VOLUME          = {287},
       YEAR            = {2002}

}

Some parts (if not most) of the "Basic Mathematical Properties" section are on the "philosophy" behind the design and use of EAs. Therefore, the author might think about changing the section title and focus on these aspects.

I recommend to remove the last paragraph from the "Basic Mathematical Properties" since it concerns a debate on the usefulness/superiority of certain versions of EAs and a rather trivial theorem known as "no free lunch theorem" not very informative for readers just looking at this field the first time. Therefore, this part should and can be reduced to one or two sentences stating that there is no single best EA. The concluding two sentences of this section summarize this simple fact and draw the right conclusions.

I can not agree that "Coevolution" deserves a separate section, at least for the time being. The author offers a bunch of citations, however, compared to the ten thousands of papers on EAs, the coevolution branch is still very small. It could be well integrated into the "Related Areas" or the "Future Vision" section. In any case, the number of citations must be reduced here (in order to meet the requirements of a good Scholarpedia article.)

The "Related Areas" section starts to blur the differences between the different branches of EAs. While the author is partly right in that "... these variations are now mostly unimportant technically, ..." the respective "... terms such as genetic algorithms, evolution strategies, evolutionary programming, and others" are still permanently used and Scholarpedia has respective entries. From that point of view, the first paragraph should be removed. Instead the second paragraph could be enhanced by providing more details on the topics mentioned there. BTW, I would not refer to "multi-objective optimization", but to "multi-objective evolutionary optimization" in the last sentence.