[Feature Request] Combine normalizing flows and GPs for flexible nonparametric/nonstationary modelling

[mrlj-hash]

🚀 Feature Request

Incorporate research developments in using normalizing flows to transform GPs for more flexible nonparametric Bayesian regression/classification. Specifically the developments in https://arxiv.org/abs/2011.01596 and https://arxiv.org/abs/2310.18230.

Motivation

Combines the strength of normalizing flows in flexible distributional modelling with the advantages of GPs for Bayesian regression. An alternative way in to nonstationary modelling compared with deep GPs. Highly scalable when combined with variational inference.

[Balandat]

This would be nice to have - would you be willing to put together a design / draft implementation for this?