Bayesian Nonparametric Dynamical Clustering of Time Series

Name: Bayesian Nonparametric Dynamical Clustering of Time Series
Uploaded: 2026-10-22T14:10:32.675407+00:00
Duration: 559 s
Description: The method models time series clusters by switching between an unknown number of linear dynamics regimes. It uses a hierarchical Dirichlet process to define these regimes and Gaussian processes to capture variations in amplitude and temporal alignment. This allows for principled clustering of time series patterns while avoiding the creation of excessive, unnecessary clusters.

The method models time series clusters by switching between an unknown number of linear dynamics regimes. It uses a hierarchical Dirichlet process to define these regimes and Gaussian processes to capture variations in amplitude and temporal alignment. This allows for principled clustering of time series patterns while avoiding the creation of excessive, unnecessary clusters.

Open the full episode