dot-active_set_refine_woodbury: Partition-Wise Active-Set Refinement for Woodbury GEE Paths

In lgspline: Lagrangian Multiplier Smoothing Splines for Smooth Function Estimation

Partition-Wise Active-Set Refinement for Woodbury GEE Paths

Description

Uses the same add/drop active-set logic as .active_set_refine(), but solves each equality-constrained subproblem through the Woodbury-corrected Lagrangian projection rather than the plain block-diagonal projection.

Usage

.active_set_refine_woodbury(
  result,
  K,
  p_expansions,
  A,
  R_constraints,
  constraint_value_vectors,
  family,
  qp_Amat,
  qp_bvec,
  qp_meq,
  rhs_list,
  Ghalf_corrected,
  wb_sqrt,
  parallel_aga,
  parallel_matmult,
  cl,
  chunk_size,
  num_chunks,
  rem_chunks,
  tol,
  parallel_qr = FALSE,
  max_as_iter = NULL,
  initial_active_ineq = integer(0)
)

Arguments

result	List of current coefficient column vectors.
K, p_expansions	Integer dimensions.
A	Original equality constraint matrix.
R_constraints	Number of columns of A.
constraint_value_vectors	Constraint RHS list.
family	GLM family object.
qp_Amat, qp_bvec, qp_meq	Inequality specification.
rhs_list	List of partition-wise right-hand-side vectors on the original coefficient scale. Gaussian GEE passes \mathbf{X}_k^{\top}\mathbf{V}^{-1}\mathbf{y}, while the non-Gaussian Woodbury path passes the partitioned score.
Ghalf_corrected	List of corrected \mathbf{G}_{\mathrm{on},k}^{1/2} matrices.
wb_sqrt	Output of .woodbury_halfsqrt_components().
parallel_aga, parallel_matmult	Logical flags.
cl, chunk_size, num_chunks, rem_chunks	Parallel parameters.
tol	Convergence tolerance.
max_as_iter	Maximum active-set iterations (default 50).
initial_active_ineq	Integer vector of inequality columns used as the initial working set.

Value

Same structure as .active_set_refine().

lgspline documentation built on Aug. 5, 2026, 1:10 a.m.