Topological density estimation

Topological density estimation and associated evaluation code/scripts

Parameter-free density estimation. Unlike cross-validation (CV), which optimizes an L2 error estimate, TDE optimizes an estimate of the so-called unimodal category. TDE outperforms CV on qualitative measures of multimodality for highly multimodal distributions. It is faster than CV and requires no parameters: just sample data and a choice of kernel (Gaussian and Epanechnikov are supported, as are histograms).