7

I’m finishing my first paper, and am struggling with the citations a bit.

I have used a particular implementation of an algorithm. The citation of those who thought of this particular variation and implementation of the algorithm is clear, but should I also include the creators of the original idea, and not just the implementation that I use?

In the case of genetic algorithms, seems like a tower where quite a lot of people have worked on (see Wikipedia for a short history), and it isn’t really clear who put the first or most important stone. Most people cite the work of J. Holland who popularized it, but that also doesn’t really feel right, since I wouldn’t give credit to all the others. On the other hand I don’t want to spend half a page repeating the history of the idea, citing 15 papers or so.

Some options that crossed my mind:

  1. Is it enough to refer to the name of the idea (of which the detailed history is quite accessible such in the example)?
  2. A huge list of citations [5,6,7,8,...,20].
  3. Referring to a source that has a nice overview of the history.

What would you advise?

6
  • 2
    I don't think #2 is practical. If I cite a paper, I don't normally cite all of the paper's citations either, that's just bloaty. Cite anything you have directly used is how I normally go about it. If it's related, you can cite it too, but don't cite everything. – Compass Nov 25 '14 at 19:19
  • "Cite anything you have directly used!" OK i like that, if i can use that as a guideline, it makes life easier. – Sarmes Nov 25 '14 at 20:10
  • You might benefit from citing Holland's article as well. If other people in the field cite it, it is likely a major article covering the topic that should sufficiently cover general information the algorithm. So a dummy citation would be like "I used a custom implementation of Normal Potato Planting[Smith], known as Sweet Potato Planting[Doe]. It gives your article a point of connectivity with a prevalent source, and a point of connectivity with the implementation of the source you've used. – Compass Nov 25 '14 at 20:18
  • @Compass In this case, Holland's work is so foundational that it is unlikely to be an appropriate citation except when discussing the history of the field. It would be like citing Darwin rather than a textbook on the modern synthesis of evolutionary biology (though less extreme since we're talking about 1970s rather than 1850s). – jakebeal Nov 25 '14 at 20:26
  • @jakebeal You doth not cite Darwin?! Monster! It's fine to cite something that cites the original too, but I didn't realize that Holland was foundational. Popularization and foundational are slightly different, after all. Darwin was foundational to evolution, but someone else definitely popularized it. – Compass Nov 25 '14 at 20:29
7

To cite a body of work large enough to have generated textbooks or comprehensive review papers, you should cite a good recent textbook or comprehensive review paper. You may not need to, however, depending on why and how you are using the algorithm. The key is to provide the reader with everything they need to know about why you chose that algorithm and the significance of the choice.

In this case, if your work is focused on genetic algorithms and/or their applications, then you should position it within the comprehensive space of genetic algorithms work. You would then cite both the textbook/review sources and the more specific "nearby" pieces to compare it to.

In the other hand, if your work is focused on an application, it's its more just about needing some tool and you happened to find genetic algorithms promising, then you don't need the comprehensive references because that's not the point. Instead, if you chose Algorithm X for particular reasons, then you should explain why you didn't pick competitor Algorithms Y, Z, and W (citing each). If you just chose Algorithm X because you thought that probably any algorithm would likely do, then be clear about that fact and don't worry about positioning it within a larger field, because honestly you haven't.

3
  • In this paper, this method is one of many used, and i didnt really want to put a lot of emphasis on it, or spend a lot of words and space to it. But i thought that i am some how oblidged to give credit to the original idea, since obviously it wasn't mine. But if i understand well, if i just use it as a tool, and didnt modify anything about it, even just a citation to the exact implementation I used would be enough? and extra references just optional? – Sarmes Nov 25 '14 at 20:08
  • @Sarmes If you picked the method because you wanted some GA method and this one happened to be handy and plausible, then you just need to state that clearly and give the one citation. What's important to communicate is whether you're saying "Method X is the best GA method for this case" (which requires comparison) vs. "Method X is a sufficient GA method for this case" (which does not, because you are not asserting anything about other methods) – jakebeal Nov 25 '14 at 20:21
  • Although the question is more about the general problem than the example, i understand you are in to the topic. My paper is intended to be about different types of test functions to compare algorithms, but since i only have 12 pages, here it will be about describing the construction of test functions. A further work should deal more with the comparisons. – Sarmes Nov 26 '14 at 11:52
4

Assuming you are writing an article for an audience that knows what a genetic algorithm is (at least, anyone in CS and related fields), it is not so important to explain what it is, as much as the implementation details. In your shoes, I would cite the paper of the implementation.

If there is a risk of readers not knowing what a genetic algorithm is, or you are using advanced details that are not common knowledge, I think it is best to add a modern textbook, or whichever explanation you think is clearest.

3

For a general reference I usually cite two papers: original/influential and a review paper.

In the case where the idea is relatively well-known and there is no single original paper, a review paper or textbook should suffice.

Unless you talk specifically about history of a discovery, or trying to put things in a historical context, there is no need for historical reviews.

In the case of doubt think about the reader. (Or imagine that you are a reader, new to this topic.)

0

Just to add a little to the other answers (which are good): don't worry about it!

As long as you cite at least one relevant source (the one you directly used) you have a lot of leeway as to which historical papers to cite. I like to do a couple things:

  • Think about who might review your paper and cite them. You can even recommend them as a reviewer.
  • Cite an often overlooked, but good quality paper. If a paper already has hundreds of citations, it doesn't need any more. But maybe there is a nice paper with a similar result that only has a handful of citations. The authors of the less cited paper will appreciate it.
  • Cite open source citations. If I have a choice people of citing a paper in a paywalled journal or an open-access journal, I will cite the open-access one, and I will cite papers from a nonprofit publisher (like PLOS or eLife) if I can.

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.