A nonparametric method of approximating something from data by assuming that itβs close to the data distribution convolved with some kernel.

This is especially popular the target is a probability density function; Then you are working with a kernel density estimator.

To learn about:

βEffective local sample sizeβ

understand the Frechet derivative + Wiener filtering construction used to derive the optimal kernel shape in (Bernacchia and Pigolotti 2011; OβBrien et al. 2016).

### Bandwidth/kernel selection in density estimation

Bernacchia and Pigolotti (2011) has a neat hack: βself consistencyβ for simultaneous kernel and distribution inference, i.e. simultaneous deconvolution and bandwidth selection. The idea is removing bias by using simple spectral methods, thereby estimating a kernel which in a certain sense would generate the data that you just observed. The results look similar to finite-sample corrections for Gaussian scale parameter estimates, but are not quite Gaussian.

Question: could it work with mixture models too?

### Mixture models

Where the number of kernels does *not* grow as fast as the number of data points,
this becomes a mixture model;
Or, if youβd like, kernel density estimates are a limiting case of
mixture model estimates.

They are so clearly similar that I think it best we not make them both feel awkward by dithering about where the free parameters are. Anyway, they are filed separately. (Battey and Liu 2013; van de Geer 1996; Zeevi and Meir 1997)discuss some useful things common to various convex combination estimators.

### Does this work with uncertain point locations?

The fact we can write the kernel density estimate as an integral with a convolution of Dirac deltas immediately suggests that we could write it as a convolution of something else, such as Gaussians. Can we recover well-behaved estimates in that case? This would be a kind of hierarchical model, possibly a typical Bayesian one.

### Does this work with asymmetric kernels?

Almost all the kernel estimates Iβve seen require KDEs to be symmetric,
because of Clineβs argument that asymmetric kernels are inadmissible
in the class of all (possibly multivariate) densities.
Presumably this implies
\(\mathcal{C}_1\) distributions,
i.e. once-differentiable ones.
In particular admissible kernels are those
which have βnonnegative Fourier transforms bounded by 1β,
which implies symmetry about the axis.
If we have an *a priori* constrained class of densities,
this might not apply.

### Fast Gauss Transform and Fast multipole methods

How to make these methods computationally feasible at scale. See Fast Gauss Transform and other related fast multipole methods.

## References

*The Annals of Statistics*6 (4): 701β26.

*Annals of the Institute of Statistical Mathematics*61 (4): 929β48.

*Case Studies in Spatial Point Process Modeling*, edited by Adrian Baddeley, Pablo Gregori, Jorge Mateu, Radu Stoica, and Dietrich Stoyan, 23β74. Lecture Notes in Statistics 185. Springer New York.

*Journal of the Royal Statistical Society: Series B (Statistical Methodology)*67 (5): 617β66.

*Nature*324 (6096): 446β49.

*Computational Statistics & Data Analysis*36 (3): 279β98.

*arXiv:1308.3968 [Stat]*, August.

*Journal of the Royal Statistical Society. Series B (Methodological)*51 (1): 81β92.

*Journal of the Royal Statistical Society: Series B (Statistical Methodology)*73 (3): 407β22.

*The Annals of Statistics*38 (5): 2916β57.

*The Annals of Statistics*16 (4): 1421β27.

*Bernoulli*20 (4): 1879β929.

*Scandinavian Journal of Statistics*35 (2): 322β34.

*Stochastic Environmental Research and Risk Assessment*27 (5): 1193β1205.

*Journal of the Royal Statistical Society. Series C (Applied Statistics)*34 (2): 138β47.

*Biometrics*35 (1): 87β101.

*Journal of the Royal Statistical Society: Series B (Statistical Methodology)*78 (2): 445β62.

*Stochastic Processes and Their Applications*39 (2): 345β58.

*Journal of the American Statistical Association*109 (505): 346β58.

*Journal of Nonparametric Statistics*6 (4): 293β310.

*SIAM Journal on Scientific and Statistical Computing*12 (1): 79β94.

*The Annals of Statistics*15 (4): 1491β1519.

*The Annals of Statistics*30 (5): 1460β79.

*Journal of Multivariate Analysis*84 (1): 19β39.

*arXiv:2012.14482 [Math, Stat]*, December.

*Institute of Mathematical Statistics Lecture Notes - Monograph Series*, 359β83. Beachwood, OH: Institute of Mathematical Statistics.

*Computational Statistics & Data Analysis*53 (10): 3726β33.

*Advances in Statistical Modeling and Inference*, 613β34.

*Methodology and Computing in Applied Probability*14 (3): 567β78.

*arXiv:1209.2082 [Cs]*, September.

*Computational Statistics & Data Analysis*72 (April): 57β76.

*Journal of Multivariate Analysis*101 (4): 949β63.

*Computational Statistics & Data Analysis*101 (September): 148β60.

*Journal of the American Statistical Association*107 (499): 1085β95.

*Biometrics*, 226β42.

*The Annals of Statistics*10 (3): 795β810.

*Neural Computation*17 (1): 19β45.

*Journal of the American Statistical Association*100 (469): 310β21.

*Advances in Neural Information Processing Systems*, 1561β68.

*Proceedings of the Ninth IEEE International Conference on Computer Vision - Volume 2*, 464β64. ICCV β03. Washington, DC, USA: IEEE Computer Society.

*Neural Networks: The Official Journal of the International Neural Network Society*10 (1): 99β109.

*Journal of Nonparametric Statistics*22 (1): 81β104.

## No comments yet. Why not leave one?