Anthony Blaom, PhD
Anthony Blaom, PhD
@jeremiedb Immensely appreciate this effort, and the feedback. Currently swamped with reviews, but I will get to this eventually. I'm guessing what you have is already can be safely assumed...
There is a [plan](https://github.com/alan-turing-institute/MLJ.jl/issues/60) and there is an undocumented method stub `update_data` in MLJModelInterface and a trait `supports_online`. I could provide guidance for a POC. We'd probably mark developments in...
> A treatment similar to what catboost/lightgbm does would then be applied. Can you point out a document explaining how these algorithms handle "categorical" features?
Thanks @pgagarinov. Cross referenced here: https://github.com/alan-turing-institute/MLJModels.jl/issues/375
@davidanthoff Any chance you could take a look at this?
I'm afraid DecisionTree.jl does not support multi-targets.
I do not know of any such plans. If you have the expertise and would like to contribute a PR to add that feature, I think we can find someone...
Thanks for reporting. That high prediction time is intriguing. Could you please provide enough detail here for others to reproduce the results, using publicly available data, or synthetic data?
This looks like progress to me. Do you think we could get away with marking the proposed change to the `Leaf` struct as non-breaking? As far as I can tell,...
Yes, but if DecisionTree.jl is using the fact that strings are ordered (they have the lexographical order) as @tlienart suggests, then DecisionTree is presumably not applying the standard splitting criterion...