I am looking at the old version (GaussianProcess) and not the new developmental version (GaussianProcessRegressor). Currently switching between 0.17 and 0.18-dev. In both cases the same eqn set for GaussianProcess is used.
I did not look at the new version as I wish to use universal kriging and I'm not sure the new version has this capability. Z ________________________________ From: Michael Eickenberg [[email protected]] Sent: Wednesday, January 06, 2016 7:09 PM To: [email protected] Subject: Re: [Scikit-learn-general] Gaussian process predict method issue Yes, it is crucial to look at your sklearn version, since @jmetzen rewrote the whole module, keeping very strictly to the algorithms detailed in Rasmussen & Williams 2006. Michael On Wednesday, January 6, 2016, Gilles Louppe <[email protected]<mailto:[email protected]>> wrote: Hi, Before going further, what version of scikit-learn are you using? We did a major update of the GP code in 0.18-dev. Best, Gilles On 6 January 2016 at 05:01, Zafer Leylek <[email protected]<UrlBlockedError.aspx>> wrote: > Just going over the scikit GaussianProcess code and comparing the results > with a simple 1D example using Rasmussen as a reference. Results are not > matching and I traced the source to the mean equation in scikit. Scikit > states the mean eqn to be: > > y_mean = f(X)^T beta + r(X)^T gamma > > but as far as I can see from Rasmussen (eqn 2.41 and 2.42) this should be > (in scikit notation): > > y_mean = -u(X)^T beta + r(X)^T gamma > > Everything else seems to match. Am I missing something here? > > Z > > ------------------------------------------------------------------------------ > > _______________________________________________ > Scikit-learn-general mailing list > [email protected]<UrlBlockedError.aspx> > https://lists.sourceforge.net/lists/listinfo/scikit-learn-general > ------------------------------------------------------------------------------ _______________________________________________ Scikit-learn-general mailing list [email protected]<UrlBlockedError.aspx> https://lists.sourceforge.net/lists/listinfo/scikit-learn-general
------------------------------------------------------------------------------
_______________________________________________ Scikit-learn-general mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/scikit-learn-general
