Adrien Corenflos

Can you explain how this algorithm is different from the auxiliary perspective of MALA in https://rss.onlinelibrary.wiley.com/doi/full/10.1111/rssb.12269? Edit: Gaussian convolution and then Langevin "denoising" is exactly aMALA in my books, so...

The marginal latent Gaussian is the counterpart for Gaussian priors, but I really think it's related: the auxiliary target is the same one, and then it looks like it's just...

But Eq (12) is immediately the gradient of the conditional proximal density though, exactly as would happen in auxiliary MALA with a different choice of decomposition in terms of $\alpha,...

No there's not yet I don't think but it's a relatively good and easy addition though we may want to be a bit careful about the design. So, my suggestion...

@junpenglao I am a bit dubious we want this algo in the library *exactly as it is*: at the end of the day it's just MALA for an auxiliary model...

I seem to not be the only one thinking the link to proximal samplers is pretty tight. 

> Hi @slinderman, would it be possible to create a new release that replaces the `tensorflow_probability` dependency with `tfp-nightly`? I think this would fix the issue for now and make...