Discover Optimal Anchors from Training Data — discover_optimal

Comprehensive anchor discovery pipeline that evaluates multiple selection strategies and quality metrics on a training dataset.

Usage

discover_optimal_anchors(
  training_data,
  n_components = 20,
  n_anchors = 100,
  n_pilot = min(10, length(training_data)),
  methods = c("mra", "coverage", "stability"),
  evaluate_stability = TRUE,
  save_results = TRUE,
  output_dir = "hatsa_anchors",
  verbose = TRUE
)

Arguments

training_data: List of training subject data matrices
n_components: Number of spectral components to use
n_anchors: Target number of anchors
n_pilot: Number of pilot subjects for initial selection (default: 10)
methods: Character vector of selection methods to try
evaluate_stability: Whether to evaluate anchor stability
save_results: Whether to save results to file
output_dir: Directory for saving results
verbose: Print progress messages

Value

An anchor_selection_result object containing: - selected_anchors: Final selected anchor indices - quality_metrics: Comprehensive quality assessment - selection_report: Detailed report of selection process