Doubly Outlier-Robust Online Infinite Hidden Markov Model

Collection of machine-learning research (arXiv, 2026-04-17) covering robustness, privacy, interpretability, and scalable inference. Key contributions: Batched Robust iHMM (BR-iHMM) — an online infinite HMM update with bounded posterior influence for outlier-robust streaming and large reductions in one-step forecasting error; Differentially Private Conformal Prediction (DPCP) — methods (differential CP and DPCP) to perform conformal prediction under differential privacy with private quantile calibration; an expert-guided, interpretable class-conditional goodness-of-fit framework applied to seim