Upon submitting your to a conference, it is recommended to attach codes (more precisely the link to the codes) for a third party to the . Ideally, you should organize the codes in a way such that one can readily verify every in your using your codes. However, this may be a bit too onerous if you compare dozens of models over several experiments. Especially, the readers may not be interested in reproducibility of intermediate results, and what really matters is, in my opinion, the performance of your best model and the baseline. Now, I have a few questions:

  1. Am I correct that it usually suffices to organize the codes in a way such that one can easily verify the result for the baseline and the best model?

  2. Do you have any tips for codes?

  3. I’ve seen many papers being scored low for their methods being unclear. In general, does attaching codes lower the chance for your paper to be scored low for this reason?

  4. I heard some conferences are trying to mandate to attach codes for reproducibility. Is this real?

Source link
thanks you RSS link
( https://www.reddit.com/r//comments/92siaz/d___for_attaching_codes_to_replicate/)


Please enter your comment!
Please enter your name here