Filtering via Simulation: Auxiliary Particle Filters

By Michael K. Pitt, N. Shephard. 1999

Subfields: Probabilistic & Bayesian Methods

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Probabilistic & Bayesian Methods

Citations in the corpus

Sequential Kernel Herding: Frank-Wolfe Optimization for Particle Filtering (2015, methodology): “…the accuracy, and (2) this type of “complicated” observation models will typically not allow for adaptation of the proposal distribution used within the ﬁlter, in the spirit of Pitt and Shephard (1999), leaving us with the standard— but ineﬃcient— bootstrap proposal as the only viable option.”
A tutorial on particle filters for online nonlinear/non-Gaussian Bayesian tracking (2002, methodology): “Although unnecessary, the original ASIR filter as proposed in [34] consisted of one more step, namely, a resampling stage, to produce an i.i.d. sample with equal weights.”
Sigma-point kalman filters for probabilistic inference in dynamic state-space models (2004, background): “It needs to be emphasized that more general proposals, which modify previously simulated trajectories, might be necessary in some scenarios [154].”
An Overview of Existing Methods and Recent Advances in Sequential Monte Carlo (2007, methodology): “The first exponents of these ideas were probably Pitt and Shephard [45], and the ideas link in closely with the biased sampling approaches proposed by [46].”
Particle Markov chain Monte Carlo methods (2010, methodology): “…it is usually recommended to set qθ.xn|yn, xn−1/ = pθ.xn|yn, xn−1/ whenever possible and to select qθ.xn|yn, xn−1/ as close as possible to pθ.xn|yn, xn−1/ otherwise; see for example Carpenter et al. (1999), Cappé et al. (2005), Doucet and Johansen (2009), Liu (2001) and Pitt and Shephard (1999).”
Snow water equivalent prediction using Bayesian data assimilation methods (2011, background): “While the SIR scheme is simple to implement, it can be inefficient, sensitive to outliers, and can quickly suffer from sample impoverishment due to the resampling at every time step (Arulampalam et al. 2002; Pitt and Shephard 1999).”
A Tutorial on Particle Filtering and Smoothing: Fifteen years later (2008, methodology): “It was originally introduced in [ 33 ] using auxiliary variables | hence its name.”
Dynamic Detection of Change Points in Long Time Series (2007, background): “While many variants exist and may be more efficient (see, for instance, Pitt and Shephard, 1999 ), the algorithm above, initially proposed by Gordon et al. (1993), typically works well for filtering a state space model whose hidden Markov process mixes well.”
Rao-Blackwellised Particle Filtering for Dynamic Bayesian Networks (2000, background): “Examples of these selection schemes include multinomial sampling (Doucet 1998, Gordon et al. 1993, Pitt and Shephard 1999), residual resampling (Kitagawa 1996, Liu and Chen 1998) and stratified sampling (Kitagawa 1996).”
Bayesian Filtering and Smoothing (2013, background): “Another variation of sequential importance resampling is the auxiliary SIR (ASIR) filter (Pitt and Shephard, 1999).”
Efficient high-dimensional importance sampling (2007, methodology): “Pitt and Shephard (1999) - produces a sequence of filtered samplers, whereby t is drawn conditionally on t 1 and y(t) only.”
Fully Automatic Facial Action Unit Detection and Temporal Analysis (2006, methodology): “Pffl is an extension to the Auxiliary Particle Filtering theory introduced by Pitt and Shephard [ 13 ], which in turn is an extension to classical particle filtering (Condensation) [7].”
The Unscented Particle Filter (2000, background): “This situation does indeed arise in several areas of engineering and finance, where one can encounter sensors that are very accurate (peaked likelihoods) or data that undergoes sudden changes (nonstationarities ): see for example (Pitt and Shephard 1999, Thrun 2000).”
Bayesian Inference and Learning in Gaussian Process State-Space Models with Particle MCMC (2013, methodology): “We discuss marginalization of f in Section 3.1 and present the inference algorithms Sections 3.2 and 3.3.”
A real-time algorithm for mobile robot mapping with applications to multi-robot and 3D mapping (2000, methodology): “The idea of using samples goes back to Rubin’s importance sampler [14] and, in the context of temporal posterior estimation is known as particle filters[12].”