diff options
Diffstat (limited to 'man/path.sparsestep.Rd')
| -rw-r--r-- | man/path.sparsestep.Rd | 44 |
1 files changed, 22 insertions, 22 deletions
diff --git a/man/path.sparsestep.Rd b/man/path.sparsestep.Rd index 935d17b..70faa7d 100644 --- a/man/path.sparsestep.Rd +++ b/man/path.sparsestep.Rd @@ -1,4 +1,4 @@ -% Generated by roxygen2 (4.1.1): do not edit by hand +% Generated by roxygen2: do not edit by hand % Please edit documentation in R/path.sparsestep.R \name{path.sparsestep} \alias{path.sparsestep} @@ -20,22 +20,22 @@ path.sparsestep(x, y, max.depth = 10, gamma0 = 1000, gammastop = 1e-04, \item{gammastop}{stopping value of the gamma parameter} -\item{IMsteps}{number of steps of the majorization algorithm to perform for +\item{IMsteps}{number of steps of the majorization algorithm to perform for each value of gamma} \item{gammastep}{factor to decrease gamma with at each step} -\item{normalize}{if TRUE, each variable is standardized to have unit L2 +\item{normalize}{if TRUE, each variable is standardized to have unit L2 norm, otherwise it is left alone.} -\item{intercept}{if TRUE, an intercept is included in the model (and not +\item{intercept}{if TRUE, an intercept is included in the model (and not penalized), otherwise no intercept is included} -\item{force.zero}{if TRUE, absolute coefficients smaller than the provided -threshold value are set to absolute zero as a post-processing step, +\item{force.zero}{if TRUE, absolute coefficients smaller than the provided +threshold value are set to absolute zero as a post-processing step, otherwise no thresholding is performed} -\item{threshold}{threshold value to use for setting coefficients to +\item{threshold}{threshold value to use for setting coefficients to absolute zero} \item{XX}{The X'X matrix; useful for repeated runs where X'X stays the same} @@ -47,7 +47,7 @@ absolute zero} \item{use.Xy}{whether or not to compute X'y and return it} } \value{ -A "sparsestep" S3 object is returned, for which print, predict, +A "sparsestep" S3 object is returned, for which print, predict, coef, and plot methods exist. It has the following items: \item{call}{The call that was used to construct the model.} \item{lambda}{The value(s) of lambda used to construct the model.} @@ -55,12 +55,12 @@ coef, and plot methods exist. It has the following items: \item{gammastop}{The gammastop value of the model} \item{IMsteps}{The IMsteps value of the model} \item{gammastep}{The gammastep value of the model} -\item{intercept}{Boolean indicating if an intercept was fitted in the +\item{intercept}{Boolean indicating if an intercept was fitted in the model} -\item{force.zero}{Boolean indicating if a force zero-setting was +\item{force.zero}{Boolean indicating if a force zero-setting was performed.} \item{threshold}{The threshold used for a forced zero-setting} -\item{beta}{The resulting coefficients stored in a sparse matrix format +\item{beta}{The resulting coefficients stored in a sparse matrix format (dgCMatrix). This matrix has dimensions nvar x nlambda} \item{a0}{The intercept vector for each value of gamma of length nlambda} \item{normx}{Vector used to normalize the columns of x} @@ -69,18 +69,18 @@ performed.} \item{Xy}{The matrix X'y if use.Xy was set to TRUE} } \description{ -Fits the entire regularization path for SparseStep using a -Golden Section search. Note that this algorithm is approximate, there is no -guarantee that the solutions _between_ induced values of lambdas do not -differ from those calculated. For instance, if solutions are calculated at -\eqn{\lambda_{i}}{\lambda[i]} and \eqn{\lambda_{i+1}}{\lambda[i+1]}, this -algorithm ensures that \eqn{\lambda_{i+1}}{\lambda[i+1]} has one more zero -than the solution at \eqn{\lambda_{i}}{\lambda[i]} (provided the recursion -depth is large enough). There is however no guarantee that there are no -different solutions between \eqn{\lambda_{i}}{\lambda[i]} and +Fits the entire regularization path for SparseStep using a +Golden Section search. Note that this algorithm is approximate, there is no +guarantee that the solutions _between_ induced values of lambdas do not +differ from those calculated. For instance, if solutions are calculated at +\eqn{\lambda_{i}}{\lambda[i]} and \eqn{\lambda_{i+1}}{\lambda[i+1]}, this +algorithm ensures that \eqn{\lambda_{i+1}}{\lambda[i+1]} has one more zero +than the solution at \eqn{\lambda_{i}}{\lambda[i]} (provided the recursion +depth is large enough). There is however no guarantee that there are no +different solutions between \eqn{\lambda_{i}}{\lambda[i]} and \eqn{\lambda_{i+1}}{\lambda[i+1]}. This is an ongoing research topic. -Note that this path algorithm is not faster than running the +Note that this path algorithm is not faster than running the \code{sparsestep} function with the same \eqn{\lambda} sequence. } \examples{ @@ -100,7 +100,7 @@ Van den Burg, G.J.J., Groenen, P.J.F. and Alfons, A. (2017). URL \url{https://arxiv.org/abs/1701.06967}. } \seealso{ -\code{\link{coef}}, \code{\link{print}}, \code{\link{predict}}, +\code{\link{coef}}, \code{\link{print}}, \code{\link{predict}}, \code{\link{plot}}, and \code{\link{sparsestep}}. } |