Specify a DKGE mapper strategy for the transport pipeline

Creates a mapper specification for the **transport pipeline** — used when mapping subject-level contrast/loading vectors from a source feature space (e.g., parcel embeddings) to a common reference space. Pass the result to [dkge_transport_spec()], [dkge_prepare_transport()], or directly to [fit_mapper()] together with `source_feat` / `target_feat` matrices.

Use [dkge_mapper()] instead when you need a **dense rendering / anchor mapper** that works in 3-D spatial coordinates (kNN barycentric, Sinkhorn OT over point clouds) for functions such as [dkge_build_renderer()] or [dkge_render_subject_values()].

Usage

dkge_mapper_spec(type = c("sinkhorn", "ridge", "ols"), ..., name = NULL)

Arguments

type: Mapping strategy identifier: `"sinkhorn"` (optimal transport), `"ridge"` (ridge regression), or `"ols"` (ordinary least squares).
...: Strategy-specific hyperparameters stored in the specification (e.g. `lambda`, `epsilon`, `lambda_emb`, `lambda_spa`).
name: Optional user-facing name for diagnostics.

Value

A `dkge_mapper_spec` object consumed by [fit_mapper()] and [predict_mapper()].

Examples

spec <- dkge_mapper_spec(type = "ridge", lambda = 1e-2)
source_feat <- matrix(rnorm(10 * 2), 10, 2)
target_feat <- matrix(rnorm(8 * 2), 8, 2)
mapping <- fit_mapper(spec, source_feat = source_feat, target_feat = target_feat)
mapped <- predict_mapper(mapping, rnorm(10))
length(mapped)
#> [1] 8