Doubly Outlier-Robust Online Infinite Hidden Markov Model

This document is an arXiv STAT/ML feed (multiple new submissions) summarizing recent machine-learning and statistical-methods papers. Key contributions include: BR-iHMM, a batched robust online infinite HMM with bounded posterior influence and improved one-step forecasting in noisy streaming settings; Differentially Private Conformal Prediction (DPCP) that combines DP model training with a private quantile mechanism to produce tighter private prediction sets; an expert-guided, interpretable class-conditional goodness-of-fit framework applied to seismic monitoring (CTBT screening); a scalable S