Partially project a new sample onto subspace — partial

Project a selected subset of column indices (colind) of new_data onto the subspace defined by the model x. Optionally do a ridge-regularized least-squares solve if columns are non-orthonormal.

Usage

partial_project(x, new_data, colind, least_squares = TRUE, lambda = 1e-06, ...)

Arguments

x: The fitted model, e.g. bi_projector, that has a partial_project method.
new_data: A numeric matrix (n x length(colind)) or vector, representing the observations to be projected.
colind: A numeric vector of column indices in the original data space that correspond to new_data's columns.
least_squares: Logical; if TRUE (default), do a ridge-regularized solve.
lambda: Numeric; ridge penalty (default 1e-6). Ignored if least_squares=FALSE.
...: Additional arguments passed to class-specific partial_project methods.

Value

A numeric matrix (n x d) of factor scores in the model's subspace, for those columns only.