datarust 0.6.0

Scikit-learn-style preprocessing and classical ML in Rust
Documentation 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229

//! Integration tests for `cluster::KMeans`.
//!
//! Mirrors the sklearn-parity testing style used elsewhere in the crate:
//! deterministic datasets with known structure, exact label recovery on
//! well-separated blobs, and round-trip serialization under the `serde` feature.

#![cfg(feature = "serde")]

use datarust::cluster::{KMeans, KMeansInit};
use datarust::serialize::{from_json, to_json};
use datarust::traits::Clusterer;
use datarust::Matrix;

fn approx(a: f64, b: f64, tol: f64) -> bool {
    (a - b).abs() < tol
}

/// Build `n_blobs` clusters of `points_per_blob` points each, each blob a tight
/// cloud around a center on the integer grid.
fn make_blobs(n_blobs: usize, points_per_blob: usize, spread: f64) -> (Matrix, Vec<usize>) {
    let mut rows: Vec<Vec<f64>> = Vec::new();
    let mut labels: Vec<usize> = Vec::new();
    let mut k = 0;
    for blob in 0..n_blobs {
        let center = (blob as f64) * 20.0; // well-separated: 20 units apart
        for i in 0..points_per_blob {
            let _ = i;
            for (dx, dy) in [(spread, 0.0), (-spread, 0.0), (0.0, spread), (0.0, -spread)] {
                rows.push(vec![center + dx, center + dy]);
                labels.push(blob);
                k += 1;
                if k >= n_blobs * points_per_blob {
                    break;
                }
            }
            if k >= n_blobs * points_per_blob {
                break;
            }
        }
    }
    (Matrix::new(rows).unwrap(), labels)
}

#[test]
fn recovers_three_blobs_up_to_label_permutation() {
    let (x, true_labels) = make_blobs(3, 12, 0.1);
    let mut km = KMeans::new()
        .with_n_clusters(3)
        .with_n_init(10)
        .with_random_state(42);
    let pred = km.fit_predict(&x).unwrap();

    // Map each predicted label to the true label of its first member.
    let mut mapping: std::collections::HashMap<usize, usize> = std::collections::HashMap::new();
    for (p, t) in pred.iter().zip(true_labels.iter()) {
        mapping.entry(*p).or_insert(*t);
    }
    // Every predicted label must map consistently to one true label.
    let correct = pred
        .iter()
        .zip(true_labels.iter())
        .filter(|(p, t)| mapping.get(p) == Some(t))
        .count();
    assert_eq!(
        correct,
        pred.len(),
        "cluster labels do not match truth (up to permutation): {:?}",
        pred
    );
}

#[test]
fn centroids_near_true_centers() {
    let (x, _) = make_blobs(3, 20, 0.05);
    let mut km = KMeans::new()
        .with_n_clusters(3)
        .with_n_init(10)
        .with_random_state(0);
    km.fit(&x).unwrap();
    let true_centers = [0.0, 20.0, 40.0];
    let mut matched = 0;
    for center in km.cluster_centers() {
        let close = true_centers
            .iter()
            .any(|&tc| approx(center[0], tc, 0.5) && approx(center[1], tc, 0.5));
        if close {
            matched += 1;
        }
    }
    assert_eq!(matched, 3, "centroids: {:?}", km.cluster_centers());
}

#[test]
fn predict_new_points_matches_nearest_centroid() {
    let (x, _) = make_blobs(2, 8, 0.1);
    let mut km = KMeans::new()
        .with_n_clusters(2)
        .with_n_init(5)
        .with_random_state(1);
    km.fit(&x).unwrap();
    // Points near 0 and near 20 should land in different clusters.
    let test = Matrix::new(vec![vec![0.0, 0.0], vec![20.0, 20.0]]).unwrap();
    let pred = km.predict(&test).unwrap();
    assert_ne!(pred[0], pred[1]);
}

#[test]
fn deterministic_with_same_seed() {
    let (x, _) = make_blobs(3, 10, 0.2);
    let mut a = KMeans::new().with_n_clusters(3).with_random_state(7);
    let mut b = KMeans::new().with_n_clusters(3).with_random_state(7);
    let la = a.fit_predict(&x).unwrap();
    let lb = b.fit_predict(&x).unwrap();
    assert_eq!(la, lb);
    assert_eq!(a.cluster_centers(), b.cluster_centers());
    assert!(approx(a.inertia(), b.inertia(), 1e-12));
}

#[test]
fn inertia_non_negative_and_lower_with_more_clusters() {
    let (x, _) = make_blobs(3, 15, 0.3);
    let mut km3 = KMeans::new().with_n_clusters(3).with_random_state(0);
    km3.fit(&x).unwrap();
    let mut km5 = KMeans::new().with_n_clusters(5).with_random_state(0);
    km5.fit(&x).unwrap();
    assert!(km3.inertia() >= 0.0);
    // More clusters can only reduce (or match) the inertia on the training set.
    assert!(km5.inertia() <= km3.inertia() + 1e-9);
}

#[test]
fn n_clusters_zero_rejected() {
    let x = Matrix::new(vec![vec![1.0], vec![2.0]]).unwrap();
    let mut km = KMeans::new().with_n_clusters(0);
    assert!(km.fit(&x).is_err());
}

#[test]
fn n_clusters_exceeds_samples_rejected() {
    let x = Matrix::new(vec![vec![1.0], vec![2.0], vec![3.0]]).unwrap();
    let mut km = KMeans::new().with_n_clusters(10);
    assert!(km.fit(&x).is_err());
}

#[test]
fn predict_before_fit_errors() {
    let km = KMeans::new().with_n_clusters(2);
    let x = Matrix::new(vec![vec![1.0], vec![2.0]]).unwrap();
    let err = km.predict(&x).unwrap_err();
    assert!(matches!(err, datarust::DatarustError::NotFitted(_)));
}

#[test]
fn predict_feature_mismatch_errors() {
    let x = Matrix::new(vec![vec![1.0, 2.0], vec![3.0, 4.0]]).unwrap();
    let mut km = KMeans::new().with_n_clusters(2).with_random_state(0);
    km.fit(&x).unwrap();
    let wrong = Matrix::new(vec![vec![1.0], vec![2.0]]).unwrap();
    let err = km.predict(&wrong).unwrap_err();
    assert!(matches!(err, datarust::DatarustError::ShapeMismatch { .. }));
}

#[test]
fn kmeans_plus_plus_finds_clusters() {
    let (x, _) = make_blobs(3, 15, 0.1);
    let mut km = KMeans::new()
        .with_n_clusters(3)
        .with_init(KMeansInit::KMeansPlusPlus)
        .with_random_state(0);
    let pred = km.fit_predict(&x).unwrap();
    let unique: std::collections::BTreeSet<usize> = pred.iter().copied().collect();
    assert_eq!(unique.len(), 3);
}

#[test]
fn random_init_finds_clusters() {
    let (x, _) = make_blobs(3, 15, 0.1);
    let mut km = KMeans::new()
        .with_n_clusters(3)
        .with_init(KMeansInit::Random)
        .with_random_state(0);
    let pred = km.fit_predict(&x).unwrap();
    let unique: std::collections::BTreeSet<usize> = pred.iter().copied().collect();
    assert_eq!(unique.len(), 3);
}

#[test]
fn n_iter_at_least_one() {
    let (x, _) = make_blobs(2, 10, 0.1);
    let mut km = KMeans::new().with_n_clusters(2).with_random_state(0);
    km.fit(&x).unwrap();
    assert!(km.n_iter() >= 1);
}

#[test]
fn serde_round_trip_preserves_predictions() {
    let (x, _) = make_blobs(3, 12, 0.1);
    let mut km = KMeans::new().with_n_clusters(3).with_random_state(42);
    km.fit(&x).unwrap();
    let before = km.predict(&x).unwrap();

    let json = to_json(&km).unwrap();
    let restored: KMeans = from_json(&json).unwrap();
    let after = restored.predict(&x).unwrap();

    assert_eq!(before, after);
    assert!(approx(restored.inertia(), km.inertia(), 1e-12));
    assert!(restored.is_fitted());
}

#[test]
fn serde_preserves_cluster_centers() {
    let (x, _) = make_blobs(2, 10, 0.1);
    let mut km = KMeans::new().with_n_clusters(2).with_random_state(0);
    km.fit(&x).unwrap();

    let json = to_json(&km).unwrap();
    let restored: KMeans = from_json(&json).unwrap();

    for (a, b) in km
        .cluster_centers()
        .iter()
        .zip(restored.cluster_centers().iter())
    {
        for (va, vb) in a.iter().zip(b.iter()) {
            assert!(approx(*va, *vb, 1e-12));
        }
    }
}