Compute a partial singular-value decomposition.

Usage

svd_partial(
  A,
  rank,
  target = largest(),
  method = auto(),
  tol = 1e-08,
  vectors = c("both", "left", "right", "none"),
  seed = NULL,
  certify = TRUE,
  allow_dense_fallback = c("auto", "never", "always")
)

Arguments

A: Matrix or eigencore operator.
rank: Number of singular values to compute.
target: Eigencore singular-value target descriptor.
method: Solver method descriptor.
tol: Convergence and certification tolerance.
vectors: Which singular-vector sides to compute.
seed: Optional random seed for stochastic solver components.
certify: Whether to compute certification diagnostics.
allow_dense_fallback: Dense fallback policy.

Value

An eigencore_svd_result containing singular values, optional left and right singular vectors, certificate diagnostics, method/plan metadata, and convergence diagnostics.

Examples

set.seed(1)
X <- matrix(rnorm(60), 10, 6)
fit <- svd_partial(X, rank = 3)
values(fit)
#> [1] 4.728358 3.042304 2.415933
certificate(fit)$passed
#> [1] TRUE