Fast & Accurate Gaussian Kernel Density Estimation

Jeffrey Heer

doi:10.1109/VIS49827.2021.9623323

UW Interactive Data Lab papers

Fast & Accurate Gaussian Kernel Density Estimation

Jeffrey Heer. IEEE VIS Short Papers, 2021

Jeffrey Heer

IEEE VIS Short Papers, 2021

Figure for Fast & Accurate Gaussian Kernel Density Estimation — Kernel density estimation error for 2D car data. Error (on a log scale) is measured as the maximum pixel error given a 100-pixel plot height. Dashed gray lines indicate the NRD bandwidth value. With 512 bins and linear binning, the Deriche method results in sub-pixel accuracy at all sampled bandwidths.

Materials

PDF | Software | Benchmarks

Abstract

Kernel density estimation (KDE) models a discrete sample of data as a continuous distribution, supporting the construction of visualizations such as violin plots, heatmaps, and contour plots. This paper draws on the statistics and image processing literature to survey efficient and scalable density estimation techniques for the common case of Gaussian kernel functions. We evaluate the accuracy and running time of these methods across multiple visualization contexts and find that the combination of linear binning and a recursive filter approximation by Deriche efficiently produces pixel-perfect estimates across a compelling range of kernel bandwidths.

BibTeX

@inproceedings{2021-fast-kde,
  title = {Fast \& Accurate Gaussian Kernel Density Estimation},
  author = {Heer, Jeffrey},
  booktitle = {IEEE VIS Short Papers},
  year = {2021},
  url = {https://uwdata.github.io/papers/fast-kde},
  doi = {10.1109/VIS49827.2021.9623323}
}