# Strange issues fitting GMMs in Turing.jl by extending a fairly simple example

**URL:** <https://discourse.julialang.org/t/strange-issues-fitting-gmms-in-turing-jl-by-extending-a-fairly-simple-example/55232>\
**Category:** Probabilistic Programming\
**Tags:** turing, bayesian-inference\
**Created:** [February 13, 2021, 10:28pm UTC](https://discourse.julialang.org/t/strange-issues-fitting-gmms-in-turing-jl-by-extending-a-fairly-simple-example/55232 "2021-02-13T22:28:45Z")\
**Posts on this page:** 1\
**Showing post:** 3

<div class="post-metadata">

**Author:** ![Christopher\_Fisher](https://sea2.discourse-cdn.com/julialang/user_avatar/discourse.julialang.org/christopher_fisher/32/26132_2.png) [@Christopher\_Fisher](https://discourse.julialang.org/u/Christopher_Fisher)\
**Post date:** [February 14, 2021, 12:42am UTC](https://discourse.julialang.org/t/strange-issues-fitting-gmms-in-turing-jl-by-extending-a-fairly-simple-example/55232/3 "2021-02-14T00:42:09Z")

</div>

There might be a few issues here. One potential issue is that the posterior density is concentrated in a small area when sigma is small, making it hard for the sampler to find high density areas. This might be exacerbated by performance issues with PG. You might try reparameterization or initializing the sampler in high density areas. In addition, marginalization might help. See this other [thread](https://discourse.julialang.org/t/mixture-models-with-binary-indicator-variables/55219).

---

_[View the full topic](https://discourse.julialang.org/t/strange-issues-fitting-gmms-in-turing-jl-by-extending-a-fairly-simple-example/55232)._