1

A high-cited paper in a field of Computer Science (that I'm working with) about an important algorithm does not tell readers about its implementation details. Since the authors were from a company, this is understandable.

In my investigation, I found that some implementation details were very important to get right, in order for my implementation to work. The overall process of getting it right took a little less than 2 months. This means that the paper could be, in general, hard for people to reproduce. There is also no open source implementation available.

My question is, is it fair to comment in a literature review that this paper "does not include some implementation details, which we have found to be highly important"? Also, does offering the implementation details in our paper count as a (maybe small) contribution?

Thanks in advance :)

2
  • Is this literature review part of a "Related Work" section / subsection in your paper? Or is it part of a more extensive review for, say, a thesis?
    – user136193
    Commented Aug 7, 2021 at 11:42
  • The literature review is part of a "Related Work" section of the paper. Commented Aug 7, 2021 at 12:09

1 Answer 1

2

Yes, you can comment on the difficulty of producing an implementation and hence, on the difficulty of reproducing the results. That is pretty natural.

But, "algorithms" as properly understood in CS don't actually depend on implementations, having a high level definition that can be implemented in several ways. That assumes, of course, that the description is correct and complete.

I can't comment on whether two months is a reasonable time to build an implementation from its abstract definition or not. It might be, of course, since implementations are often imperfect.

To turn it around, an implementation (specific code) is often the worst way to define/describe an algorithm since it may depend on language/machine specifics that are not inherent in the algorithm itself. The code may not give the same insight that a higher level description could provide. That may well be the reasoning of the originators, not an attempt to protect trade secrets or related things.

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .