| Title: | The Chopthin Resampler |
|---|---|
| Description: | Resampling is a standard step in particle filtering and in sequential Monte Carlo. This package implements the chopthin resampler, which keeps a bound on the ratio between the largest and the smallest weights after resampling. |
| Authors: | Axel Gandy and F. Din-Houn Lau |
| Maintainer: | Axel Gandy <[email protected]> |
| License: | GPL-3 |
| Version: | 0.2.2 |
| Built: | 2024-11-20 03:49:37 UTC |
| Source: | https://github.com/cran/chopthin |
A fast implementation of the Chopthin resampler. Can be used as the resampling step in particle filters and in sequential Monte Carlo.
chopthin(w, N, eta = 5.828427, normalise = TRUE, checks = TRUE)chopthin(w, N, eta = 5.828427, normalise = TRUE, checks = TRUE)
w |
a vector of weights |
N |
target number of particles |
eta |
upper bound on the ratio between the weights. Must be >=4. If eta=Inf then only thinnig is performed, requiring the number of weights to be at least N. |
normalise |
Flag for controlling if the returned weights should be normalised. If TRUE (default) then the sum of the returned weights will sum to N. If FALSE then the returned weights have the same sum as the original weights (within the numerical precision). |
checks |
Flag controlling if checks on the input and the result should be performed. Default TRUE. |
A list with two elements: new weights and indices of the ancestors of the new particles. The weights are normalised to add up to N.
A Gandy and F. D-H Lau. The chopthin algorithm for resampling. IEEE Transactions on Signal Processing, 64(16):4273–4281, 2016
chopthin(runif(10),N=10) chopthin(runif(10),N=20,4) chopthin(runif(10),N=5) chopthin(runif(10),N=1)chopthin(runif(10),N=10) chopthin(runif(10),N=20,4) chopthin(runif(10),N=5) chopthin(runif(10),N=1)