The nested particle filter of Crisan and Miguez (2018)

[Sahel13]

Reference - Crisan and Miguez (2018)

The main thing to decide is how we want to deal with the parameters. At the moment I have added function types that have an explicit param argument (see cuthbert/npf/types.py).

TODOs

• The jittering kernel accepts kwargs, but nothing is passed to it in filter_combine
• Adaptive jittering (jitter only after resampling) - This is problematic, we currently have no way of reliably knowing if resampling has happened or not. The only way I can think of is to have resampling functions return an additional boolean output.
• More sophisticated tests
• General documentation and an example

[Sahel13]

I'd like to get your opinions on having an explicit param argument @SamDuffield @AdrienCorenflos

[SamDuffield]

Sorry I'm thinking I need to understand the algorithm better. Ideally we can do without explicit param handling, i.e. by just running a PF on a modified SSM on an enlarged space. But I need to understand better if that makes sense for this algorithm