Rémi Louf
Rémi Louf
I will make it easier to decouple sampling and updating parameters in #276 . This should simplify the code a lot.
CC @pierrejacob
https://arxiv.org/abs/2104.05134
https://arxiv.org/abs/2206.07202
@MarcoGorelli what could be causing that?
@AdrienCorenflos Alright I rebased your branch on the main branch (fixed many merge conflicts) and squashed the commits. Two things before merging: - I don't like the kwarg and branching...
I think I've figured out what bothered me. The base SMC kernel was designed as a mere sampler when it is in fact much more general. Reading your book more...
> So what you're saying by "more general" than a "mere sampler" This statement is about _my_ former mental representation of SMC, which comes from existing probabilistic languages. They tend...
I've never been a big fan on this API to be honest, I expect it to return the parameters every time I use it. So maybe we should do just...
Linked to #171. Should be addressed together.