It's OK in general, but you have to check if the journal accepts it
May I ask you, in the field of deep learning, if I find uneven distribution of other people's data sets in the review of others' achievements, I will further select their data sets, and then get better model performance. Is this a fair comparison with the original author, or is it an improvement on the original author's work?
0 Answer
It's OK in general, but you have to check if the journal accepts it
Yes, all get more optimized models, of course there are improvements, can not say a fair comparison, you are on the basis of pressing deep development, the results must be improved,
If you are reproducing someone else's work and you find that the original author's data set is unevenly distributed, and you have further filtered and processed it for better performance, Then you can compare your results to the original authors. In this case, however, you need to document very clearly what you did with the data so that others can reproduce your work and understand your results.
You can explain why your performance is better than the original author's by presenting and demonstrating the advantages of your data processing strategy. Doing so increases the credibility of your results and helps convince others that you are justified in your results.
In summary, if you have adopted a better strategy for handling the data than the original author and can demonstrate the superiority of that strategy, then your results can be considered an improvement on the original author's work.
Reply given by AI
这家伙很懒，什么都没留下...
