Talk:Matérn covariance function

Statistics Low‑importance

	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 articles
Low	This article has been rated as Low-importance on the importance scale.

Mathematics Low‑priority

	Mathematics portal This article is within the scope of WikiProject Mathematics, a collaborative effort to improve the coverage of mathematics 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.MathematicsWikipedia:WikiProject MathematicsTemplate:WikiProject Mathematicsmathematics articles
Low	This article has been rated as Low-priority on the project's priority scale.

Yes but why?[edit]

This (stub) article basically says what a Matérn covariance function is, and the purposes for which it is used. But for someone mathematical who has never heard of it, there's no text at all about why use Matérn covariance function rather than radial basis function, Student's T, or any of the common functions one can easily use to build a stationary kernel? What is it about Matérn that makes it good for geospatial stats? Can someone please put an answer into the article? --mcld (talk) 22:25, 13 January 2014 (UTC)[reply]

What is it that makes the logit link good for binary regression? Tradition, I'm afraid. -- Anonymous — Preceding unsigned comment added by 129.67.26.155 (talk) 09:25, 10 February 2014 (UTC)[reply]

Continuity claim[edit]

The article currently says a Gaussian process with Matern covariance is floor(v-1) times differentiable. According to [1] (p 85), it is "k-times MS differentiable if and only if ν > k". Indeed, for v = 3/2, a GP is once-differentiable. Perhaps I misread something, but it seems floor(v-1) should be replaced by floor(v)$. Can anyone confirm this? Ksimek (talk) 17:20, 23 July 2015 (UTC)[reply]

[1] C. Williams and C. Rasmussen, Gaussian processes for machine learning. MIT Press, 2006.

In its present form, the claim is that it is ceil(ν)-1 times differentiable. For ν = 3/2, this is the same as floor(ν), but it differs for integers. When ν = 1, ceil(ν)-1 = 0, but floor(ν) = 1. In this case the behavior of ceil(ν)-1 matches the statement given in Williams and Rasmussen, so we can't just replace it with floor(ν). Nejssor (talk) 03:14, 6 February 2020 (UTC)[reply]

"Matern kernel" listed at Redirects for discussion[edit]

An editor has asked for a discussion to address the redirect Matern kernel. Please participate in the redirect discussion if you wish to do so. Utopes _{(talk / cont)} 02:15, 12 April 2020 (UTC)[reply]