Closed Marcdh3 closed 2 years ago
That does indeed seem like a bug. But from what I can tell, it was your Nov 2019 commit that introduced this code--so my question would be, what reference were you using?
1-dist becomes a similarity metric, so my guess is we're using the metric down the road for spectral analysis of an affinity matrix. But it would be good to double-check this is actually happening.
git blame
can't follow that history.
I have been comparing our implementation of the Hellinger distance for two multivariate normal distributions with the equation defined on Wikipedia, and I am left with two questions: