BR-iHMM adds bounded-PIF robustness to online iHMMs via generalized Bayesian updates and two extra tuning parameters, cutting forecasting error by up to 67% on three real and synthetic streams.
We assume the outlier is the first observation of the batch for convenience, the location of the outlier within the batch is irrelevant
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
stat.ML 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Doubly Outlier-Robust Online Infinite Hidden Markov Model
BR-iHMM adds bounded-PIF robustness to online iHMMs via generalized Bayesian updates and two extra tuning parameters, cutting forecasting error by up to 67% on three real and synthetic streams.