Closed devmotion closed 3 years ago
Thanks! It's important to keep old code compatible with newer versions of the libraries. However, if you are just starting with NB, consider using MLJ.jl, which contains an implementation originating from this package, but actively maintained.
Basically, MLJ wraps this package, so you might run into issues anyway. In this case opening tickets or contributing to this package is indeed the best strategy.
Ah I hadn't checked MLJ. I thought for my simple toy example it would be sufficient to use something more lightweight. Afterwards I noticed that NaiveBayes already has quite many dependencies (I was particularly surprised about HDF5), so it doesn't seem particularly lightweight either.
In addition to this PR, it might be good to add CompatHelper to the repo.