# Blackjax / Contribute !

###

* We need documentation *now*.
* It would be nice if the documentation also acts like a literature review. Ie: give the pros/cons of each algorithm and tuning recommendations. Similar to [sgmcmcjax's sampler page](https://sgmcmcjax.readthedocs.io/en/latest/all_samplers.html) or to this [list of MCMC algorithms](https://m-clark.github.io/docs/ld_mcmc/index_onepage.html). The idea is that BlackJAX could attract users even just with the documentation/lit review.

## Examples

* Any comparison with optimization
* Comparisons between sampling algorithm
* Anything with many dimensions / lots of datapoints
* Anything that explains how the algorithms work

## Algorithms

- Likelihood-free inference, in particular one-shot SMC ABC
- Riemannian HMC (implicit and explicit integration)
- Slice sampling
- Variational inference:
- Mean field approximation
- [SVGD](https://arxiv.org/abs/1608.04471)
- Some kind of "ensemble" method (note that ensemble can mean many different things), such as [Ensemble preconditing MCMC](https://arxiv.org/abs/1607.03954)

## Adaptation

- ChEEs
- Empirical HMC

## Meta
- Gibbs sampling
- HMC coupling
- Inference loops
- Parallel tempering

## Testing

- [Simulation Based Calibration](https://arxiv.org/pdf/1804.06788.pdf)
- [Kernel goodness-of-fit test](http://proceedings.mlr.press/v48/chwialkowski16.pdf)