1.6.0

• Plotting Mode: Native, high-quality visualization engine that generates publication-ready stacked bar charts. It features greedy maximum-overlap cluster alignment across runs to keep colors consistent, and supports up to three hierarchical tiers of population label brackets alongside custom color palettes.
• Shared SVD for Fast Multi-K Sweeps: Significantly accelerated parameter sweeps across ranges of $K$ by computing and reusing a single randomized SVD initialization once, avoiding redundant calculations and heavily boosting training speeds.
• Cross-Validation: Built-in v-fold cross-validation (--cv) that automatically masks genotype entries and measures prediction error, providing a mathematical and robust way to identify the optimal number of ancestral clusters.
• Supervised Mode: Enables reference-guided clustering by clamping the ancestry proportions ($Q$) of known reference samples, anchoring the allele-frequency model ($P$) to direct genetic backgrounds while estimating the remaining samples freely.
• Projection Mode: Allows rapid ancestry inference ($Q$-only optimization) for new target datasets against pre-trained, fixed reference panels ($P$) without the need to retrain or modify the baseline model.