Several new algorithms are presented every day, but when somebody is trying to reproduce outputs then he has to 'build' code in R, FORTRAN, Matlab etc by himself.
I think that this is one strong reason for repulsion from the main idea presented at the published paper.
What if I decide to include a simple example of R or FORTRAN code in order to be 'ready for use' from the scientific community?
Could this downgrade the theoretical contribution of the algorithm?
What is your opinion?
(As I Know Econometrica is a Journal requiring reproducibility of the results with sending the codes that have been used for a submitted article, but codes are not being published)
Reproductibility is a key point in scientific publications. Thus, if you are presenting some numerical results in your paper, a good practice is to make your instances available (unless explicitly forbidden by the company you're working in/with). You can still go even further by publishing a working sample of your code, but this is optional. You do not need to put the code in the paper, pointing to a permanent webpage where one can download the instances/code is sufficient. Then, I cannot guaranty your paper will be accepted, but you are adding some good points for your paper.
Besides, I don't think adding some code snippets reduces the scientific contribution of a paper, whether it is theoretical or practical.
Adding code in the manuscript is not good idea, rather you can provide the link the site from which the code can be downloaded.
adding full code into a paper is not a very good idea. What i suggest maybe you can put the algorithm of how the method works then it is still can reproduce by using any programing languange. Some time we need also to concern our property right and at the same time our research contributions....
Nazri: so your opinion is that a paper should be written in such a way that any researcher could reproduce results by working with his favorite 'tool' easily?
Code presented in a print or PDF format is not particularly useful, and no one sits and *reads* computer code. If the algorithm is short enough, pseudocode in the article will do; but otherwise, I recommend making it available to download on a third-party site like, for instance, R-forge. Putting it on your own website will be less permanent; if you move hosts, URLs may change, making the link in the paper dead (this has happened to me).
Echoing the other responses, if you publish a paper in which you make a claim using computer code, the claims should be reproducible, preferably *exactly*. This is your responsibility as an author of a scientific journal article, and it requires releasing your code.
I just wanted to add that many journals are now accepting "online supplements", which could give authors a wonderful incentive to upload their code and data sets in a unified package and host it as a supplement on the journal's website.
The only downside is that sometimes, the supplement may be behind a paywall, but if people are not familiar with online software repositories, it may be the next best thing. Upload the main article, images, references, etc. and the supplementary stuff and be done with it in one step.
One additional point to these: I would absolutely want the code described in your publication published in a readily accessible place. If not on the publisher website, many sites provide ways to propagate your code in a public manner, while limiting the change authority: Mathworks, octave-forge, R-forge, source-forge, CodeProject and many others provide project oriented or domain oriented publication, freely accessible for no cost to the author.
So, till now, we have:
3 YES
3 NO
2 MAYBE
I think there is not a clear direction at the 'debate' yet.
I think you can provide the link of the code only. But make sure that it can be downloadable
While I don't see it as an absolute must-have in any algorithms-related article, presenting a sample code is very useful for practical applications of your research, and particularly, I wouldn't complain if I were to find articles on algorithms that happen to have a good implementation example.
You have to keep in mind, though, that the form with which you'll present the program in the article must be, shall I say, elegant enough that it "blends in" with the rest of the paper, and adds to it.
EDIT: Well, now that I think of it, I'm with the people that are suggesting an online URL from which you can download a sample code.
A printed (or PDF) code that cannot be easily C/P'd into a text editor and compiled/interpreted right away is much less practical than an electronic format from which you can download and run directly.
Just in case, maybe you should present both the code and a download link in your article. None of them are absolutely necessary, but they are useful.
- Badges
- Science topic
- Similar topics
- Computer Science and Engineering
- Algorithms