Ryan Curtin

That's correct, we'll need to cache the query trees and modify them also. There's no need to extract the reference trees from the KNN objects, you can just store the...

Right, imagine that we have some `KDTree queryTree`. Then we'll need to write something a little like: ``` // First we need to stretch the dataset. queryTree.Dataset() = transformationMatrix *...

> I did a quick derivation, the change in points is basically arma::mat change = ((Transformation - TransformationOld) * dataset) Right, exactly, and we can see our hyperrectangle bound as...

Oops, forgot one: > new minWidth should be basically the max(node.Bound[d].Hi() - node.Bound[d].Lo()) Shouldn't that be `min()` not `max()`?

So you are right that the dimension with maximum variance can indeed change. But that should not affect the correctness of the results (it may cause the search to be...

So, definitely when you open the PR we should include the test that you wrote. A couple of notes, first about your test directly above: when you call `knn.Train(dataset)`, the...

Any update on this one?

Can you push the code that you have? I'd like to try and debug this. :)

Hey @david-cortes, thanks for the report. I opened a PR, #3057, that at least fixes the documentation associated with logistic regression. Actually, this is a little difficult to fix, because...

Honestly, I am not an expert at R (I can barely write it), so if you can tell me what would be more typical in R, that could be really...