Prepare reusable rendering objects for a fitted DKGE model

Usage

dkge_build_renderer(
  fit,
  centroids,
  anchors = NULL,
  anchor_xyz = NULL,
  anchor_n = 20000L,
  anchor_method = c("kmeans", "sample"),
  anchor_seed = NULL,
  vox_xyz = NULL,
  mapper = dkge_mapper("knn", k = 8, sigx = 3),
  graph_k = NULL,
  decoder_k = 8,
  reliabilities = NULL,
  subject_feats = NULL,
  anchor_feats = NULL,
  feat_lambda = NULL,
  feat_sigma = NULL
)

Arguments

fit: Fitted `dkge` object.
centroids: List of per-subject centroid matrices (`P_s x 3`). Must align with `fit$Btil`.
anchors: Optional precomputed anchor coordinate matrix. When `NULL`, anchors are derived from `anchor_xyz` if provided, otherwise from `vox_xyz`.
anchor_xyz: Optional matrix of candidate points used to derive anchors via [dkge_make_anchors()]. Ignored when `anchors` is supplied.
anchor_n: Number of anchors to draw when constructing them from coordinates.
anchor_method: Method passed to [dkge_make_anchors()] when anchors are derived. Defaults to `"kmeans"`.
anchor_seed: Optional seed forwarded to [dkge_make_anchors()].
vox_xyz: Optional voxel coordinates for constructing a decoder.
mapper: Mapper specification created with [dkge_mapper()]. Defaults to the barycentric kNN mapper.
graph_k: Optional integer; when provided, an anchor graph of this neighbourhood size is constructed for subsequent smoothing.
decoder_k: Number of anchors per voxel when building the decoder.
reliabilities: Optional list of per-subject reliability vectors passed to the mapper during fitting.
subject_feats: Optional list of matrices supplying latent features per subject cluster. When provided and the mapper consumes latent information (e.g., Sinkhorn), they are forwarded via `subj_feats`.
anchor_feats: Optional anchor-level feature matrix aligned with `anchors`. Derived automatically by pooling subject features when `NULL` and `subject_feats` are provided.
feat_lambda: Feature cost weight passed to Sinkhorn mappers. Ignored by kNN.
feat_sigma: Feature bandwidth used when computing feature costs.

Value

A list bundling anchors, optional graph/decoder, fitted per-subject mappers, and subject weights.

Examples

# \donttest{
toy <- dkge_sim_toy(
  factors = list(A = list(L = 2), B = list(L = 3)),
  active_terms = c("A", "B"), S = 3, P = 15, snr = 5
)
fit <- dkge(toy$B_list, toy$X_list, kernel = toy$K, rank = 2)
#> Warning: Argument 'kernel' is deprecated; use 'K' instead.
centroids <- lapply(toy$B_list, function(B) matrix(rnorm(ncol(B) * 3), ncol(B), 3))
renderer <- dkge_build_renderer(fit,
                                centroids = centroids,
                                anchor_xyz = matrix(rnorm(20 * 3), 20, 3),
                                anchor_n = 20,
                                anchor_method = "sample")
length(renderer$anchors)
#> [1] 60
# }