Publications Google Scholar

Quantifying the Stability–Recovery–Interpretability Trade-Off Between K-Means and Self-Organizing Maps for High-Dimensional Imbalanced Data NEW

Quantifies how partition stability, minority class recovery, and topological interpretability trade off across K-means, SOMs, consensus K-means, and four modern baselines on 10,095 TCGA cancer RNA samples (13,634 genes, 19 cancer types). At matched prototype budgets, K-means outperforms SOMs on both stability and minority recovery, while SOMs uniquely provide 2D topological visualization. No single method optimizes all three properties.