KDE implementation for synthetic data of planetary systems. Here, I determine the optimal bandwidth for multivariate probability density function through KernelDensity (sklearn) and gaussian_kde (scipy). To plot 3d pdfs is used plotly library.
Launch it in binder:
The python code is the same notebook, but without plots. It just has the multivariate case, in order to run it in a cluster.
Some usefull infromation of Kernel density estimation (KDE) and the bandwith problem in slides