Skip to contents

Fit PARROT on a pair of domains

Source: R/adapters_parrot.R
fit_pair.parrot_aligner.Rd

PARROT-specific method for pairwise alignment using optimal transport with random walk regularization. Returns a soft assignment/transport plan.

Usage

# S3 method for class 'parrot_aligner'
fit_pair(
  algo,
  X_i,
  X_j,
  links = NULL,
  ncomp = NULL,
  sigma = 0.15,
  lambda = 0.1,
  lambda_e = NULL,
  lambda_n = NULL,
  lambda_p = NULL,
  tau = 0.05,
  alpha = 0.2,
  gamma = 0.1,
  solver = c("sinkhorn"),
  max_iter = 100,
  tol = 1e-06,
  use_cpp = FALSE,
  anchors_policy = c("warn", "off", "error"),
  unsup_strategy = c("mnn_bootstrap", "attribute_only", "self_training"),
  conf_threshold = 0.9,
  max_rounds = 1,
  seed_strength = 0.05,
  laplacian_reg = NULL,
  ...
)

Arguments

algo

A parrot_aligner object

X_i

First domain data matrix (samples x features)

X_j

Second domain data matrix (samples x features)

Optional correspondence links; list(vec1, vec2) with label vectors or NA for unmatched samples

ncomp

Number of spectral components (currently unused)

sigma

RWR restart probability

lambda

Overall regularization weight

lambda_e

Edge/attribute cost weight (defaults to lambda * 0.5)

lambda_n

Network/Laplacian regularization (defaults to lambda * 0.5)

lambda_p

Anchor prior weight (defaults to lambda)

tau

Entropic regularization parameter for Sinkhorn

alpha

Mixing parameter for cost matrix

gamma

Network structure penalty weight

solver

Transport solver method ("sinkhorn")

max_iter

Maximum iterations for RWR and Sinkhorn

tol

Convergence tolerance

use_cpp

Logical; use C++ implementation if available

anchors_policy

Policy when no anchors provided: "warn", "off", "error"

unsup_strategy

Unsupervised seeding strategy: "mnn_bootstrap", "attribute_only", or "self_training"

conf_threshold

Confidence threshold for self-training

max_rounds

Maximum rounds for iterative strategies

seed_strength

Anchor prior strength when using MNN seeds

laplacian_reg

Alternative to lambda_n for Laplacian regularization

...

Additional arguments (unused)

Value

An object of class parrot_pair_fit containing transport (n1 x n2 matrix), objective (final cost), n1 (samples in domain i), and n2 (samples in domain j).

Examples

# \donttest{
set.seed(456)
X1 <- matrix(rnorm(80), 40, 2)
X2 <- matrix(rnorm(80), 40, 2)
algo <- parrot_aligner()
fit <- fit_pair(algo, X1, X2, sigma = 0.2, tau = 0.1)
#> Warning: PARROT: no anchors provided; using unsupervised seeding (mnn_bootstrap)
class(fit)
#> [1] "parrot_pair_fit"
# }