Infinite hidden Markov models for cylindrical data

2026-07-08

AI Summary

This work addresses the challenge of modeling cylindrical time series when the number of underlying states is unknown and the joint distribution of angular and radial components is complex. The authors propose a Bayesian nonparametric infinite hidden Markov model, extending this framework for the first time to cylindrical data. The model employs a conjugate von Mises–Gamma emission distribution to jointly characterize angular and radial features and leverages a beam sampler for efficient posterior inference. Experimental results demonstrate that the proposed method achieves superior modeling accuracy and adaptive capacity on both synthetic data and two real-world datasets, effectively capturing the intrinsic structure of cylindrical time series without requiring prespecification of the state count.

Abstract

We propose an infinite hidden Markov model for cylindrical time series with von Mises-Gamma emissions. Posterior inference is performed using a beam sampler combining conjugate updates and approximate sampling schemes. Simulation studies and two real data applications demonstrate the effectiveness of the proposed methodology.

Problem

• infinite hidden Markov models
• cylindrical data
• von Mises-Gamma emissions
• time series

Innovation

• infinite hidden Markov model
• cylindrical data
• von Mises-Gamma emissions
• beam sampler
• nonparametric Bayesian