Bayesian Inference for Linear Dynamic Models with Dirichlet Process Mixtures

Abstract

Using Kalman techniques, it is possible to perform optimal estimation in linear Gaussian state-space models. We address here the case where the noise probability density functions are of unknown functional form. A flexible Bayesian nonparametric noise model based on Dirichlet process mixtures is introduced. Efficient Markov chain Monte Carlo and Sequential Monte Carlo methods are then developed to perform optimal batch and sequential estimation in such contexts. The algorithms are applied to blind deconvolution and change point detection. Experimental results on synthetic and real data demonstrate the efficiency of this approach in various contexts.

Cited by in corpus (13)

• Bayesian Nonparametric Inference of Switching Linear Dynamical Systems
• Bayesian semiparametric inference for multivariate doubly-interval-censored data
• Dynamic density estimation with diffusive Dirichlet mixtures
• Bayesian nonparametric forecasting of monotonic functional time series
• Conjugacy properties of time-evolving Dirichlet and gamma random measures
• Centered Partition Process: Informative Priors for Clustering
• Sequential Dirichlet Process Mixtures of Multivariate Skew t-distributions for Model-based Clustering of Flow Cytometry Data
• Use of Bayesian Nonparametric methods for Estimating the Measurements in High Clutter
• Bayesian Nonparametric Density Autoregression with Lag Selection
• Graph Gamma Process Generalized Linear Dynamical Systems
• Robust Particle Filtering via Bayesian Nonparametric Outlier Modeling
• Bayesian Inference in Nonparametric Dynamic State-Space Models
• A Bayesian Nonparametric Markovian Model for Nonstationary Time Series