sfpca_cd_solve_cpp: Coordinate descent for the SFPCA penalized quadratic...

In genpca: Generalized Principal Component Analysis

Coordinate descent for the SFPCA penalized quadratic subproblem

Description

Internal solver for ⁠min_x 0.5 x'Sx - b'x + P(x; lambda)⁠ with sparse SPD S and an L1 or SCAD penalty.

Usage

sfpca_cd_solve_cpp(S, b, x0, lambda, penalty, scad_a, max_sweeps, tol)

Arguments

S	sparse SPD matrix (dgCMatrix)
b	numeric vector, linear term
x0	numeric vector, warm start
lambda	penalty level (must be >= 0)
penalty	0 for L1, 1 for SCAD
scad_a	SCAD shape parameter (> 2)
max_sweeps	maximum number of full-equivalent sweeps
tol	convergence tolerance on the KKT residual (gradient units)

Value

list with x, sweeps, and kkt (max KKT residual)

genpca documentation built on Sept. 17, 2026, 1:09 a.m.