Control settings for `gpca_align.hyperdesign()`

Usage

gpca_align_control(
  knn = NA_integer_,
  knn_mode = c("symmetric", "mutual"),
  balance = c("none", "within", "all"),
  balance_power = 1,
  normalize = c("fro", "edges", "degree"),
  verbose = TRUE,
  max_dense_elems = 2e+08,
  memory_fraction = 0.5,
  memory_safety_bytes = 512 * 1024^2
)

Arguments

knn: Optional integer k for k-NN sparsification (applied to within and between graphs)
knn_mode: Whether to use a "symmetric" (union) or "mutual" (intersection) k-NN graph
balance: Whether to apply label balancing to "none", "within", or "all" edges
balance_power: Exponent for inverse-frequency label scaling (>= 0)
normalize: Normalisation mode for the component metrics: "fro", "edges", or "degree"
verbose: Logical toggle for informational messages
max_dense_elems: Maximum dense elements permitted when forming the block matrix (set `NA` to auto-tune)
memory_fraction: Fraction of available RAM to consider safe when auto-tuning densification
memory_safety_bytes: Absolute byte cushion to subtract when auto-tuning densification

Value

A list of class `gpca_align_control`

Examples

ctrl <- gpca_align_control(knn = 15, normalize = "fro")