scholarly journals Particle Filtering, Learning, and Smoothing for Mixed-Frequency State-Space Models

2016 ◽  
Author(s):  
Markus Leippold ◽  
Hanlin Yang
Keyword(s):  
State Space ◽  
Particle Filtering ◽  
State Space Models ◽  
Mixed Frequency ◽  
Frequency State

Approximate Conditional Mean Particle Filtering for Linear/Nonlinear Dynamic State Space Models

2008 ◽  
Vol 56 (12) ◽  
pp. 5790-5803 ◽  
Author(s):  
Derek Yee ◽  
J.P. Reilly ◽  
T. Kirubarajan ◽  
K. Punithakumar
Keyword(s):  
State Space ◽  
Particle Filtering ◽  
Nonlinear Dynamic ◽  
State Space Models ◽  
Dynamic State ◽  
Conditional Mean

scholarly journals Metropolising forward particle filtering for particle smoothers

2020 ◽  
Author(s):  
Monty Rubin ◽  
Jim Dowell ◽  
Nathan Strachan ◽  
Edwin Weintraub
Keyword(s):  
Monte Carlo ◽  
State Space ◽  
Particle Filtering ◽  
Conditional Distribution ◽  
State Space Models ◽  
State Trajectory ◽  
New Methods ◽  
Inverse Variance ◽  
Finite State ◽  
Finite State Space

Smoothing in state-space models amounts to computing the conditional distribution of the latent state trajectory, given observations,or expectations of functionals of the state trajectory with respect tothis distributions. For models that are not linear Gaussian or possess finite state space, smoothing distributions are in general infeasibleto compute as they involve intergrals over a space of dimensionalityat least equal to the number of observations. Recent years have seenan increased interest in Monte Carlo-based methods for smoothing,often involving particle filters. One such method is to approximatefilter distributions with a particle filter, and then to simulate backwards on the trellis of particles using a backward kernel. We showthat by supplementing this procedure with a Metropolis-Hastings stepdeciding whether to accept a proposed trajectory or not, one obtainsa Markov chain Monte Carlo scheme whose stationary distribution isthe exact smoothing distribution. We also show that in this procedure,backward sampling can be replaced by backward smoothing, which effectively means averaging over all possible trajectories. In an examplewe compare these approaches to a similar one recently proposed by Andrieu, Doucet and Holenstein, and show that the new methods can bemore efficient in terms of precision (inverse variance) per computationtime.

Sign in / Sign up

Export Citation Format

Share Document