Have you compared the latency and accuracy of streaming k‑means versus an online Gaussian mixture for real‑time anomaly detection?