Current internal data structure uses a circular reference, highly undesirable for future stability and code cleanliness. Probably worth replacing ball tree aspect with kNN.jl
Hi @johnmyleswhite, this package uses BallTrees to represent KDEs, but is probably not the best Julia implementation. I was considering replacing the underlying datastructure with an existing Julia one, and the search led me to your kNN.jl. Would you advise I build a kNN.jl dependency into KernelDensityEstimate.jl?
This KDE package has built-in multi-scale Gibbs sampling for multidimensional density, multi-term product approximation, and hence the original need for yet another KDE package. Maybe there is enough overlap and reason to converge the 3 Julia KDE packages?
dehann
commented
7 years ago
Current internal data structure uses a circular reference, highly undesirable for future stability and code cleanliness. Probably worth replacing ball tree aspect with kNN.jl
Hi @johnmyleswhite, this package uses BallTrees to represent KDEs, but is probably not the best Julia implementation. I was considering replacing the underlying datastructure with an existing Julia one, and the search led me to your kNN.jl. Would you advise I build a kNN.jl dependency into KernelDensityEstimate.jl?
This KDE package has built-in multi-scale Gibbs sampling for multidimensional density, multi-term product approximation, and hence the original need for yet another KDE package. Maybe there is enough overlap and reason to converge the 3 Julia KDE packages?