0
$\begingroup$

I'm working on a project where I need to compare the similarity of two probability distributions using MMD. However, the two datasets have different sample sizes. I've read that MMD can be affected by sample size differences, but I'm not sure to what extent.

Can I still use MMD for my task, even with unequal sample sizes? Are there any adjusted versions of MMD or specific kernel bandwidth selection strategies that are recommended for this scenario?

Improve this question
$\endgroup$

1 Answer 1

Reset to default
1
$\begingroup$

the two datasets have different sample sizes

It's unclear why this is even an issue for you.

Just draw N examples without replacement from each dataset, and you're off to the races.

Verify that MMD produces repeatable results on your distributions by doing this multiple times, and comparing the results.

Improve this answer
$\endgroup$

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.