kernel.pls.fit {plsdof} | R Documentation |
This function computes the Kernel Partial Least Squares fit.
kernel.pls.fit(K, y, m, compute.DoF, step.size)
K |
kernel matrix. This is a square matrix with dimension = number of observations. We assume that the data is centered in the feature space. |
y |
vector of response observations. |
m |
number of Partial Least Squares components |
compute.DoF |
Logical variable. If compute.DoF=TRUE , the Degrees of Freedom of Partial Least Squares are computed.
|
step.size |
After how many steps should the latent components be re-orthogonalized? See below for more details. |
This function computed the kernel coefficients for Partial Least Squares and its Degrees of Freedom. The computation
relies on a sparse matrix decomposition of the predictor variables. The sparse structure can sometimes lead to non-orthogonal latent components
(due to numerical problems). To avoid this, we orthogonalize the latent components after step.size
iterations.
Alpha |
matrix of kernel coefficients. This is a matrix of size length(y) x m . |
dYhat |
array of first derivatives of \widehat y with respect to y . The dimension of dYhat equals
[m,length(y),length(y)]. |
Yhat |
matrix of fitted values. This is a matrix of size length(y) x m . |
DoF |
vector of Degrees of Freedom. |
RSS |
vector of residual sum of squares |
sigmahat |
vector of estimated model error |
TT |
matrix of latent components |
Nicole Kraemer, Mikio L. Braun
Kraemer, N., Braun, M.L. (2007) "Kernelizing PLS, Degrees of Freedom, and Efficient Model Selection", Proceedings of the 24th International Conference on Machine Learning, Omni Press, 441 - 448
kernel.pls
, kernel.pls.cv
, kernel.pls.ic
## this is an internal function called by kernel.pls