I realized currently there is no library for multivariate kernel density estimation (unless you extend some existing code). And I need it for one of my projects. So here it is: MultiKDE.jl.
- Multidimension: Using unnormalized generalization product kernel to estimate multi-dimensional kernel density (refers statsmodels).
- Lazy evaluation: Doesn’t pre-initialize a KDE, only evaluate points when necessary.
- Categorical distribution: This library supports categorical KDE using two specific kernel functions, both ordered-categorical (age, amount…) and unordered-categorical (sex, face of the coin…). When using unordered-categorical, non-numeric objects are also supported.