Tailoring Density Estimation via Reproducing Kernel Moment Matching

Abstract

Moment matching is a popular means of parametric density estimation. We extend this technique to nonparametric estimation of mixture models. Our approach works by embedding distributions into a reproducing kernel Hilbert space, and performing moment matching in that space. This allows us to tailor density estimators to a function class of interest (i.e., for which we would like to compute expectations). We show our density estimation approach is useful in applications such as message compression in graphical models, and image classification and retrieval.

Cite

Text

Song et al. "Tailoring Density Estimation via Reproducing Kernel Moment Matching." International Conference on Machine Learning, 2008. doi:10.1145/1390156.1390281

Markdown

[Song et al. "Tailoring Density Estimation via Reproducing Kernel Moment Matching." International Conference on Machine Learning, 2008.](https://mlanthology.org/icml/2008/song2008icml-tailoring/) doi:10.1145/1390156.1390281

BibTeX

@inproceedings{song2008icml-tailoring,
  title     = {{Tailoring Density Estimation via Reproducing Kernel Moment Matching}},
  author    = {Song, Le and Zhang, Xinhua and Smola, Alexander J. and Gretton, Arthur and Schölkopf, Bernhard},
  booktitle = {International Conference on Machine Learning},
  year      = {2008},
  pages     = {992-999},
  doi       = {10.1145/1390156.1390281},
  url       = {https://mlanthology.org/icml/2008/song2008icml-tailoring/}
}