Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Tutorial 2: Unsupervised Learning using Bayesian Mixture Models #2175

Open
CMoebus opened this issue Feb 25, 2024 · 0 comments
Open

Tutorial 2: Unsupervised Learning using Bayesian Mixture Models #2175

CMoebus opened this issue Feb 25, 2024 · 0 comments

Comments

@CMoebus
Copy link

CMoebus commented Feb 25, 2024

Hi,
I wanted to generalize the tutorial's code a bit so that cluster 2D-cluster centroids could possess unconstrained components. In the tutorial the x- and y- components are constrained to be equal. In the tutorial this is achieved by using only a 1D-parameter vector $\mu = [-3.5, 0.5]$. I rewrote the code only a tiny bit so that these equality constraints are avoided. Now, I have replaced the 1D-vector $\mu = [-3.5, 0.5]$ by one DxK array with cluster centroids as column vectors. Despite all my code tinkering I were not successful in the identification of the two clusters. Now I want to ask you whether you could modify your tutorial code so that the equality constraint in $\mu-$components is avoided. Then it would be possible to run the algorithm with the Old Faitful data set so that a comparison with results of EM algorithm (cf. Bishop, 2006, p.437, Fig 9.8) is possible.

A second issue is that I cannot install Turing.jl v0.30.5. My configuration is Win11 and

[31c24e10] Distributions v0.25.107
[7876af07] Example v0.5.3
[1a297f60] FillArrays v1.9.3
[c3e4b0f8] Pluto v0.19.38
[f3b207a7] StatsPlots v0.15.7
[fce5fe82] Turing v0.30.4

Thanks in advance, Claus Möbus

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

1 participant