Gaussian Mixture Model estimation in Julia 1.0

Hello,

I wonder, if there are any up-to-date Julia packages for GMM estimation? I googled around, and came across https://github.com/davidavdav/GaussianMixtures.jl and https://github.com/lindahua/MixtureModels.jl. Both look outdated. Distributions.jl has support for mixture models, but has no estimation procedure.

Are there other packages that I have missed? Any suggestions on how best to proceed?

Thanks,
Evgeny

1 Like

I am not sure what you mean, as the last commit for

was 9 days ago an its tests pass on Julia v1.0.

Hmm, fair enough. But here is what I get when I try to install GaussianMixtures:

 pkg> add GaussianMixtures@0.2.0
 Resolving package versions...
ERROR: Unsatisfiable requirements detected for package GaussianMixtures [cc18c42c]:
 GaussianMixtures [cc18c42c] log:
 ├─possible versions are: [0.0.1-0.0.12, 0.1.0, 0.2.0] or uninstalled
 ├─restricted to versions 0.2.0 by an explicit requirement, leaving only versions 0.2.0
 └─restricted by julia compatibility requirements to versions: uninstalled — no versions left

Not sure what this means…

what happens if you just write

add GaussianMixtures

(or: why did you add @0.2.0 ?)

what happens if you just write …

The same thing happens. I added the version explicitly, hoping that this would fix it.

When I go

pkg> dev GaussianModels

The package gets cloned from Git, and all the dependencies get built and the example works fine. So looks like something got tripped up in the package manifest.

What’s the output of ] st?

GaussianMixtures.jl seems to have Logging.jl in its deps which shows no signs of supporting Julia 1.0, so may be this is a case of Package compatibility caps, worth filing an issue.

1 Like

The update to 1.0 was just not tagged. Use ]add GaussianMixtures#master

2 Likes

hi all,

just trying to get a sense of who is actively working on this right now. I just submitted a PR to fix the test suite and I would be interested to work on different EM implementations. could do with a little help though. anyone?

1 Like