clustering_evidence/
clustering_evidence.rs

1use sim_lib_numbers_stats::{
2    CovarianceType, GmmControl, GmmSpec, KMeansControl, SingularComponentPolicy, fit_gmm,
3    fit_kmeans,
4};
5
6fn main() {
7    let points = vec![
8        vec![-3.1, -3.0],
9        vec![-3.0, -2.9],
10        vec![-2.9, -3.1],
11        vec![4.9, 5.0],
12        vec![5.0, 5.1],
13        vec![5.1, 4.9],
14    ];
15    let kmeans = fit_kmeans(
16        &points,
17        2,
18        KMeansControl::new(4, 50, 1.0e-10, 20_000, 4).expect("control"),
19    )
20    .expect("bounded k-means");
21    let selected = &kmeans.restarts[kmeans.selected_restart];
22    println!(
23        "kmeans centroids={:?} inertia={:.6} selected={} restarts={} converged={} work={}",
24        kmeans.model.centroids,
25        selected.inertia,
26        kmeans.selected_restart,
27        kmeans.restarts.len(),
28        selected.converged,
29        kmeans.work
30    );
31
32    let gmm = fit_gmm(
33        &points,
34        GmmSpec::new(
35            2,
36            CovarianceType::Diagonal,
37            1.0e-6,
38            SingularComponentPolicy::default(),
39        )
40        .expect("spec"),
41        GmmControl::new(4, 50, 1.0e-9, 100_000).expect("control"),
42    )
43    .expect("regularized GMM");
44    println!(
45        "gmm means={:?} likelihood={:.6} iterations={} converged={} repairs={} aic={:.6} bic={:.6} work={}",
46        gmm.model.means,
47        gmm.evidence.log_likelihood,
48        gmm.evidence.iterations,
49        gmm.evidence.converged,
50        gmm.evidence.singular_component_repairs,
51        gmm.evidence.model_selection.aic,
52        gmm.evidence.model_selection.bic,
53        gmm.evidence.work
54    );
55}
clustering_evidence/clustering_evidence.rs

clustering_evidence/
clustering_evidence.rs
