Align multiple datasets by synchronizing pairwise transforms

This meta-aligner lifts any pairwise method to N domains by: 1) choosing a sparse dataset graph, 2) fitting pairwise models on edges, 3) extracting relative transforms, and 4) running group-appropriate synchronization to get per-domain transforms into a shared space.

Usage

align_many(
  domains,
  algo,
  graph = c("auto", "mst", "complete", "star"),
  consensus = c("auto", "sync", "star", "tree"),
  k = NULL,
  parallel = TRUE,
  sync = NULL,
  ...
)

Arguments

domains: list of domain objects (often matrices), one per dataset
algo: an aligner object created by new_aligner() with adapter methods
graph: "auto" | "mst" | "complete" | "star"
consensus: "auto" | "sync" | "star" | "tree" (currently: "sync"/"star")
k: target latent dimension (optional)
parallel: logical; reserved for future parallel pair fits
sync: list of synchronization options (graph, refine, gl, perm settings)
...: forwarded to adapter methods

Value

list with transforms, embeddings (if apply_transform succeeds), dataset graph, pairwise fits, and diagnostics

Details

If the adapter reports native multi-view support, dispatches to fit_many().

Examples

# \donttest{
set.seed(1)
X1 <- matrix(rnorm(50), 10, 5)
X2 <- matrix(rnorm(50), 10, 5)
X3 <- matrix(rnorm(50), 10, 5)
domains <- list(X1, X2, X3)
# align_many requires an aligner object with adapter methods
# See package vignettes for complete examples
# }