Seth Axen

This is provided now by LogExpFunctions's ChangesOfVariables extension: https://github.com/JuliaStats/LogExpFunctions.jl/blob/a1c4fda2b9cc4c59c184648c0cfc7f694c415bf3/ext/LogExpFunctionsChangesOfVariablesExt.jl#L7-L10

This is already supported with the `weights` keyword.

Thanks! This seems to work well! Here is how I separately time the entire warm-up phase and the sampling phase: ```julia function extract_initialization(state) (; Q, κ, ϵ) = state return...

It probably makes sense for these bijectors to return a `Factorization` object but for their inverses to accept not only a `Factorization` but also an iterable of the factors. e.g....

Or does it make more sense to in an extension to a UniqueFactorizations.jl package, overload the Bijectors interface for functions like `UniqueFactorizations.qr`? Is there anything in Bijectors expecting all bijectors...

@torfjelde @devmotion what do you think?

> Can you give me an example code in python-xarray which writes and reads such a Boolean array? Sure, here's an example: ```python >>> import xarray as xr >>> import...

There's also a work-in-progress paper here explaining the theory behind the method: https://github.com/aseyboldt/covadapt-paper/blob/main/main.pdf

Thanks, @aseyboldt for the pointers! I'll check back in if I get lost. Btw, I was scanning your draft of the paper, and it seems like a really nice contribution....

> The mass matrix is initialized using the gradient (https://github.com/pymc-devs/nuts-rs/blob/main/src/adapt_strategy.rs#L267) Why this particular initialization choice? Sometimes $J^TJ$ for Jacobian $J$ is used to approximate the Hessian, which would correspond to...