Does GMM have the option to get the cluster probability?

[rickbeeloo]

Hey!

After doing the fit:

// We fit the model from the dataset setting some options
    let gmm = GaussianMixtureModel::params(n_clusters)
                .n_runs(10)
                .tolerance(1e-4)
                .with_rng(rng)
                .fit(&dataset).expect("GMM fitting");

    // Then we can get dataset membership information, targets contain **cluster indexes**
    // corresponding to the cluster infos in the list of GMM means and covariances
    let blobs_dataset = gmm.predict(dataset);

Can I get the probability of it belonging to one of the clusters? It will assign it to the cluster with the highest probability (I assume) however, I want to only assign if the probability is higher than a specific threshold.

I'm not familiar with linfa at all, so perhaps there is a standard way of doing this.

Thanks for building all this!

[relf]

Hi. Unfortunatly, at the moment, this information is not made available.

The good news is that the implementation is rather straightforward. If I understand correctly you need predict_proba() as implemented in scikit-learn. As linfa GMM implementation is a direct port of the scikit-learn one, the code is pretty similar and the method can be easily ported. Let me know if you want to open a PR on this.

[techyorzl]

Hi @relf , I’d like to work on this issue. I’m currently learning Rust, but I’ve gone through the codebase and believe I can solve this. Could you assign it to me?