# Discretized HMC

**URL:** <https://discourse.julialang.org/t/discretized-hmc/59964>\
**Category:** Probabilistic Programming\
**Created:** [April 25, 2021, 2:16am UTC](https://discourse.julialang.org/t/discretized-hmc/59964 "2021-04-25T02:16:50Z")\
**Posts on this page:** 2\
**Page:** 1

<div class="post-metadata">

**Author:** ![ParadaCarleton](https://sea2.discourse-cdn.com/julialang/user_avatar/discourse.julialang.org/paradacarleton/32/20005_2.png) [@ParadaCarleton](https://discourse.julialang.org/u/ParadaCarleton)\
**Post date:** [April 25, 2021, 2:16am UTC](https://discourse.julialang.org/t/discretized-hmc/59964/1 "2021-04-25T02:16:50Z")

</div>

Is there any interest in implementing generalizations of HMC to discrete spaces for Turing.jl? Examples:

[https://www.tandfonline.com/doi/full/10.1080/01621459.2019.1585255?casa\_token=XuidyAblCZQAAAAA:FrOpfFdJWbBrG1R0-ZNZ6pfgDnK15hhj6mXvDWrTbZwZCQCrm0K7SM5lDVE5P-bDUwWxK7TRrQ77NA](https://www.tandfonline.com/doi/full/10.1080/01621459.2019.1585255?casa_token=XuidyAblCZQAAAAA:FrOpfFdJWbBrG1R0-ZNZ6pfgDnK15hhj6mXvDWrTbZwZCQCrm0K7SM5lDVE5P-bDUwWxK7TRrQ77NA)

> **[Sampling in Combinatorial Spaces with SurVAE Flow Augmented MCMC](http://proceedings.mlr.press/v130/jaini21a.html)**
>
> Hybrid Monte Carlo is a powerful Markov Chain Monte Carlo method for sampling from complex continuous distributions. However, a major limitation of HMC is its inability to be applied to discrete d...

---

<div class="post-metadata">

**Author:** ![cpfiffer](https://sea2.discourse-cdn.com/julialang/user_avatar/discourse.julialang.org/cpfiffer/32/208747_2.png) [@cpfiffer](https://discourse.julialang.org/u/cpfiffer)\
**Post date:** [April 25, 2021, 2:25pm UTC](https://discourse.julialang.org/t/discretized-hmc/59964/2 "2021-04-25T14:25:48Z")

</div>

Yeah, I think we’d be interested in seeing how that would integrate into Turing – we’d love to have someone try to open a PR somewhere for this.