STPGA - Selection of Training Populations by Genetic Algorithm
Advanced genetic algorithms for optimal subset selection
in high-dimensional prediction problems. Provides efficient
single and multi-objective optimization for training population
selection with comprehensive criteria including A-, D-,
E-optimality, prediction error variance (PEV), Cook's distance
(CD), and distance-based measures. Features multi-criteria
convergence detection, restart mechanisms, rank-based selection
with pressure control, adaptive mutation, diversity
preservation, and numerically stable matrix operations.
Includes convergence diagnostics, configurable optimization
windows, and both modern clean interfaces with legacy
compatibility functions.
