Talk:Variable kernel density estimation

	This article is within the scope of WikiProject Statistics, a collaborative effort to improve the coverage of statistics on Wikipedia. If you would like to participate, please visit the project page, where you can join the discussion and see a list of open tasks.StatisticsWikipedia:WikiProject StatisticsTemplate:WikiProject StatisticsStatistics
???	This article has not yet received a rating on Wikipedia's content assessment scale.
???	This article has not yet received a rating on the importance scale.

There seems to be a factor 1/h missing in the first section called Rationale. Compare here: http://en.wikipedia.org/wiki/Kernel_density_estimation#Definition

Good call. In my own paper, the h is a sigma and it is absorbed into K while the K is not normalized and there is a separate normalization coefficient. However, I decided in this article to pull out the bandwidth, change the symbol and normalize K to bring my notation more in line with others, not realizing that you also have to pull h out of the normalization coefficient. Thanks. Peteymills (talk) 23:16, 13 May 2013 (UTC)[reply]

nested kernel estimators :

nested kernel estimators = multilayer perceptron (or feedforward neural networks)

the "multi-variate kernel density estimation" is the special case where the kernel's parameter depends on space directions, to be general in fact the kernel could depend on the point of space, where it's parameters are given by another "kernel density estimator",

this way each layer (each kernel density estimator) is the layer of a (feed-forward) neural network,

and we also include hidden markov models (for discrete random variables)

78.227.78.135 (talk) 01:10, 30 October 2015 (UTC)[reply]