19

We have developed a novel algorithm, published a paper about it, then after a few months, we "copied" our implementation to a very popular opensource library.

A few months after, a paper was published proposing an algorithm that was compared against all off-the-shelf algorithms from the above-mentioned library, including ours.

The results of our algorithm were bad and inconsistent, and when we double-checked the implementation again in the library, we found out that because the library has a certain convention, our algorithm was incompatible (it's our fault, no doubt). Please note that the results in our paper along with our "local" implementation are 100% correct and valid.

We have fixed the implementation in the library, and now it is 100% correct and similar to the algorithm that we had published in the literature.

Now we are kind of confused what to do, Should we inform the authors of the new paper who cited and used our old implementation in the library? Or Should we let it go even if it might tarnish the image of our algorithm?

33

I don't see any reason not to contact the authors of the other paper, though it is possible that nothing can be done for an already published paper. But the conversation with them might be useful to everyone.

I, personally, would thank them for alerting you to a problem with the published algorithm, giving you the opportunity/incentive to make a correction.

Possibly a note could be inserted somewhere. One simple possibility is for you to make it clear, perhaps within the repository, that the algorithm has been updated as of a certain date and that earlier implementations were flawed.

In other words, try to make it possible for a reader to be made aware of all relevant facts.

1
  • 3
    It is also possible, for some journals, to submit a "response" to a published article that criticizes one's work. I'm not sure whether this would be an option in this case, but if you feel that your good name has been impugned, it might be worth looking into. Jul 21 at 17:21

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.