0
$\begingroup$

If one would like to study aspects of neural networks (say, in an academic paper), and would like to experiment on binary classification of vectors in low-dimensional space (say dim=2 or dim<6); what are standard/good datasets for that? Probably synthetic, but possibly also not. So, I am thinking of datasets which are ``generic" enough to be considered representative for an academic study of neural net behavior.

$\endgroup$

1 Answer 1

0
$\begingroup$

There is no such thing as a generic enough dataset. If you are planning for an academic study, I would suggest to use multiple different datasets that cover a huge variety of characteristics. Most good data science studies do so.

If you are looking for real world datasets then it would be best to search repositories like kaggle, openML or UCI for datasets that suit your case.

Note: Depending on the subject of the study, one should be careful to generalize from low-dimensional datasets to general aspects of neural networks. But without knowing more about the study, I can not say whether this would be a point.

$\endgroup$
1
  • $\begingroup$ Thanks. Yes, I meant, say, a series of datasets which will be kind-of representative in some sense, so that an exposed phenomenon on it will be presumed as a "real" phenomenon which probably happens generically enough. I also thought about your point about low-dim vs. high-dim... It is just that what I check is quite computation-heavy so for now I thought checking lower-dim stuff. But I will have this in mind, thanks. $\endgroup$
    – Sasha
    Apr 23, 2023 at 17:47

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.