The hardware and bandwidth for this mirror is donated by dogado GmbH, the Webhosting and Full Service-Cloud Provider. Check out our Wordpress Tutorial.
If you wish to report a bug, or if you are interested in having us mirror your free-software or open-source project, please feel free to contact us at mirror[@]dogado.de.
This version introduces a complete overhaul of the package, with substantial improvements to both the core computational engine and the user-facing interface.
The main change in this version is the exploitation of the banded structure of the penalization matrix, which allows for tremendous savings in both computation time and memory footprint.
The package now uses Rcpp and calls directly Lapack functions for banded matrices when available.
Some additional functions have also been coded using Rcpp. The
backsolve function for banded matrix was also recoded
although it already exists in Lapack because the Lapack version
overwrites the Cholesky factor so a copy has to be made
beforehand.
Parameter estimation now rely on backward-forward solves from the Cholesky factor instead of forming the variance-covariance matrix explicitly.
By default, only the diagonal of the variance-covariance matrix
is now computed (which is sufficient for credibility intervals). The
full matrix can still be retrieved using the newly introduced
vcov() method if necessary.
The two main functions from earlier versions have been merged
into a single unified function: WH().
Performance iteration and rank reduction are no longer available, as they no longer provide meaningful benefits with the new implementation. Indeed, the full-rank version is now efficient enough to handle several thousand observation points without optimization tricks.
predict.WH_1d and predict.WH_2d functions. It
turns out that the formula used for the extrapolation ignored the
innovation error caused by the prior on the extrapolated region,
resulting in smaller credibility intervals than they should have been.
This now has been fixed.expect_equal(f(x), f(x))
after confirmation they do not work with MKL BLASFurther improved test robustness
Added tests of the form expect_equal(f(x), f(x)) for
testing purposes
Simplified computation of the matrix tUWU by using
the crossprod function, which should reduce memory usage
(by half) and computation time (slightly)
Fixed an issue with the WH_2d plot and the what =
“edf” argument
Improved tests robustness
NEWS.md file to track changes to the
packageThese binaries (installable software) and packages are in development.
They may not be fully stable and should be used with caution. We make no claims about them.
Health stats visible at Monitor.