Resampling API

This section describes the resampling API of GenericSSMs.jl and is intended for users who wish to use their own resampling algorithm with GenericSSMs.jl. Otherwise, the resampling algorithms of Resamplings.jl implement the required functionality and can be directly used with GenericSSMs.jl.

For particle filtering (pf_forward_pass!), the resampling object should implement:

resample!(resampling, ind::AbstractVector{<: Integer}, 
          w::AbstractVector{<: AbstractFloat}, rng)::Nothing

resample! should draw ancestor indices to ind given normalised weights w. Note that length(ind) == length(w)` should hold. The last argument is preserved for a random number generator.

For conditional particle filtering (cpf_forward_pass!), the resampling object should implement:

conditional_resample!(resampling, ind::AbstractVector{<: Integer}, 
                      w::AbstractVector{<: AbstractFloat}, 
                      k::Integer, i:::Integer, rng)::Nothing

conditional_resample! should draw ancestor indices to ind given normalised weights w and the condition that ind[k] = i. Again, length(ind) == length(w) should hold, and additionally, w[i] should be strictly positive. The last argument is preserved for a random number generator.

In general the resampling object can be of any type (and naturally contain arbitrary fields), as long as it satisfies the above API. It may be useful to look at the documentation and source code of Resamplings.jl for examples.