4

I have a new algorithm which I would like to publish.

I'm not a student, so I wanted to know a few things: 1) What should I publish? Just the algorithm with explanation? an academic paper? What are the principals of writing such one? 2) Where to publish? 3) What are the outcomes of publishing the algorithm? Legally or any other aspects I should be aware of.

Any other tips and helpful things I should know will be welcomed

user51503
  • 41
  • 1
  • 2

2 Answers2

5

Making an algorithm public is easy: you can put it on your webpage. By having it published as an academic paper (either conference or journal), you can get reviews, which is interesting. Reviewers may help you evaluate the novelty of your approach. I thus suggest a paper, with at least:

  • a motivation for your algorithm,
  • a review of the related approaches and the explanation for its novelty,
  • the algorithm with explanations.

It is always a plus if you implement it also in a common computer language, so that others can check its efficiency, find bugs, allow improved versions. A working version of your algorithm in the shape of a code (compilable with instructions) can be put on GitHub for instance, for better versioning.

There are several algorithm related journals and conferences, for instance at the ACM Society:

and you can easily find many more, depending on the subfield your algorithm belong (combinatorics, optimization, real-time computing, etc.). Some are listed at the List of computer science conferences: Algorithms and theory.

Laurent Duval
  • 4,995
  • 15
  • 35
  • 1
    Conferences are the primary venue for disseminating research in algorithms, not journals. See, for example, SODA. – JeffE Jan 01 '17 at 14:18
  • 1
    @JeffE True. When I wrote "published as an academic paper", I had in mind conferences too, because they often count as "papers". In computer vision, machine learning, many conferences indeed publish algorithmic papers – Laurent Duval Jan 01 '17 at 14:54
  • To @LaurentDuval - If algorithm is published as academic paper in journal (as version 1.0), is it possible to later publish its new upgraded version? (2.0, 3.0 etc.) ? Would the editors of mathematical/IT journals make a complain that it is "not novel enough" ? Do you have any experience with such scenario? – Ladislav Naďo Jan 17 '18 at 21:59
  • 1
    I cannot say for all algorithms. But this happens often in optimization, and for instance with the FFTW. Of course, editors can complain, however if the novel version is shown more efficient, with better convergence, implement in parallel, formally proven. Maybe a short note could be good enough for a V2.0 – Laurent Duval Jan 18 '18 at 20:55
1

I suggest you to go for arXiv. However, you cannot just publish a plain algorithm. You need to format your document as a research paper. This link might give you an idea of what to include.

padawan
  • 12,245
  • 8
  • 48
  • 83