View source for Talk:Crossover (evolutionary algorithm)

{{WikiProject banner shell|class=B|
{{WikiProject Robotics|importance=mid}}
{{WikiProject Molecular Biology|COMPBIO=yes|COMPBIO-importance=low}}
}}

==Untitled==
Please start numbering at 1, not 0. Starting at 0 is a programmer's constraint.

What is crossover rate?
a crossover rate is the probability that a crossover may happen --[[Special:Contributions/124.168.53.12|124.168.53.12]] ([[User talk:124.168.53.12|talk]]) 13:52, 8 April 2009 (UTC)

== Relation to Chromosomal crossover ==

How does this article relate to [[Chromosomal crossover]]? --[[User:Kvng|Kvng]] ([[User talk:Kvng|talk]]) 02:50, 2 April 2010 (UTC)

== The paragraph "Ordered Chromosomes" ==
User Ficeup has done a rather bad job on 17th of Dec. 2010 introducing this paragraph - simply deleting an existing paragraph which in my eyes was maybe not brilliant, but at least proper English. There were multiple typos and the problem is, and even worse you really don't understand what his/her point is:

: One such case is when the chromosome is an ordered list, such as an ordered list of the cities to be travelled for the [[traveling salesman problem]]. 
: There are many crossover methods for ordered chromosomes, '''of course we can also by using''' the crossover method mentioned above(N-point crossover), '''but sometimes we need a repair method to repair it'''. '''We can also by consider following methods''' (we use a simple [[travelling salesman problem]] as an example),

Does anyone understand this and is able to clarify?

Furthermore, the same user simply changed the original fixed probability value from 0.5 to 0.2. However, due to this change, there seems to be a contradiction to an earlier sentence in the same paragraph:

: The Uniform Crossover evaluates each bit in the parent strings for exchange '''with a probability of 0.5'''. [...]
: In the uniform crossover scheme (UX) individual bits in the string are compared between two parents. The bits are swapped with a fixed probability, '''typically 0.2'''.

Again, it is not clear whether this really should be 0.5 and not 0.2! [[Special:Contributions/193.134.202.252|193.134.202.252]] ([[User talk:193.134.202.252|talk]]) 09:06, 14 April 2011 (UTC) e_l_

== Uniform crossover ==

Why is uniform crossover referred to as "a poor method"? There's no explanation, justification, or sources. I'm not saying it's not true -- I just don't know -- and think that part of the article isn't clear.  <span style="font-size: smaller;" class="autosigned">— Preceding [[Wikipedia:Signatures|unsigned]] comment added by [[Special:Contributions/129.79.114.206|129.79.114.206]] ([[User talk:129.79.114.206|talk]]) 18:12, 14 June 2013 (UTC)</span><!-- Template:Unsigned IP --> <!--Autosigned by SineBot-->

I added a dubious marker, as I can't see why this would obviously be a "poor method" as it seems equally valid as any of the other methods. For that matter it appears that the other methods favor continuous runs, of "good" data. This really only makes sense if the algorithm is working on integers, not necessarily some more abstract method. One way to look at "uniform crossover" is as a "n-point crossover".

Furthermore the following sentences contradict this statement, implying that it is in fact better. The article [[Premature convergence]] also lists this as a way to prevent premature convergence.
[[User:KillerGardevoir|KillerGardevoir]] ([[User talk:KillerGardevoir|talk]]) 19:47, 6 June 2014 (UTC)

Whether or not uniform crossover is "a poor method" is completely subjective. It requires more generations to traverse the same amount of search space as other crossover operators, which means it can take longer, but that is a trade-off for a more exhaustive search. The end result is contingent on many different factors, but all other things being equal, there is no reason why you can't achieve equally low error margins with UCX as other methods. [[User:Seetomgo|Seetomgo]] ([[User talk:Seetomgo|talk]]) 22:00, 9 January 2015 (UTC)

== Removed some sections ==

Removed these things from the article:

* Methods of selection. Selection has nothing to do with crossover, they are totally disjoint components of a genetic algorithm. It has its own Wikipedia page.

* Graphic accompanying uniform crossover. The crossover it shows is not actually a uniform crossover, but a k-point crossover. It also has irrelevant text in the graphic.

* Three parent crossover. Crossover is almost always between two parents. This page is not the place for (very very very) exotic examples.

* Section "For ordered genomes". Removed the explanations for some of the crossover methods, which I thought were very unclear. It would be best if someone adds new explanations.  <!-- Template:Unsigned --><small class="autosigned">—&nbsp;Preceding [[Wikipedia:Signatures|unsigned]] comment added by [[User:Theelepeltje|Theelepeltje]] ([[User talk:Theelepeltje#top|talk]] • [[Special:Contributions/Theelepeltje|contribs]]) 13:59, 30 May 2018 (UTC)</small> <!--Autosigned by SineBot-->

== Thorough revision of the old article ==

The old article was greatly expanded, since crossover operators occur not only in [[Genetic algorithm|genetic algorithms]] but in almost all other [[Evolutionary algorithm|evolutionary algorithms]] (EA). In this respect, the title is also rather too restrictive. The design of crossover operators depends not only on the task but also very much on the representation of the decision variables of the chromosome ([[genetic representation]]). More examples were given and, above all, nuch more references were made so that it easier for the interested reader to find more information. [[User:Studi90|Studi90]] ([[User talk:Studi90|talk]]) 16:10, 16 January 2023 (UTC)