aprender-core 0.29.1

Next-generation machine learning library in pure Rust
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

use super::*;

#[test]
fn test_tsne_very_short_iterations() {
    let data = Matrix::from_vec(4, 2, vec![1.0, 2.0, 2.0, 3.0, 10.0, 11.0, 11.0, 12.0])
        .expect("valid matrix dimensions");

    // Test with very few iterations (to hit early iteration paths)
    let mut tsne = TSNE::new(2).with_n_iter(10).with_random_state(42);
    let result = tsne
        .fit_transform(&data)
        .expect("fit_transform should succeed");
    assert_eq!(result.shape(), (4, 2));
}

#[test]
fn test_tsne_high_iterations() {
    let data = Matrix::from_vec(4, 2, vec![1.0, 2.0, 2.0, 3.0, 10.0, 11.0, 11.0, 12.0])
        .expect("valid matrix dimensions");

    // Test with iterations past momentum switch (250+)
    let mut tsne = TSNE::new(2).with_n_iter(300).with_random_state(42);
    let result = tsne
        .fit_transform(&data)
        .expect("fit_transform should succeed");
    assert_eq!(result.shape(), (4, 2));
}

#[test]
fn test_tsne_single_component() {
    let data = Matrix::from_vec(
        4,
        3,
        vec![
            1.0, 2.0, 3.0, 2.0, 3.0, 4.0, 10.0, 11.0, 12.0, 11.0, 12.0, 13.0,
        ],
    )
    .expect("valid matrix dimensions");

    // Reduce to 1D
    let mut tsne = TSNE::new(1).with_n_iter(100).with_random_state(42);
    let result = tsne
        .fit_transform(&data)
        .expect("fit_transform should succeed");
    assert_eq!(result.shape(), (4, 1));
}

#[test]
fn test_tsne_very_low_perplexity() {
    let data = Matrix::from_vec(
        6,
        3,
        vec![
            1.0, 2.0, 3.0, 1.5, 2.5, 3.5, 5.0, 6.0, 7.0, 5.5, 6.5, 7.5, 10.0, 11.0, 12.0, 10.5,
            11.5, 12.5,
        ],
    )
    .expect("valid matrix dimensions");

    // Very low perplexity to test binary search extremes
    let mut tsne = TSNE::new(2)
        .with_perplexity(1.5)
        .with_n_iter(50)
        .with_random_state(42);
    let result = tsne
        .fit_transform(&data)
        .expect("fit_transform should succeed");
    assert_eq!(result.shape(), (6, 2));
}

#[test]
fn test_tsne_identical_points() {
    // Test with nearly identical points (tests numerical stability)
    let data = Matrix::from_vec(4, 2, vec![1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0])
        .expect("valid matrix dimensions");

    let mut tsne = TSNE::new(2).with_n_iter(50).with_random_state(42);
    let result = tsne
        .fit_transform(&data)
        .expect("fit_transform should succeed");
    assert_eq!(result.shape(), (4, 2));

    // Results should be finite
    for i in 0..4 {
        for j in 0..2 {
            assert!(result.get(i, j).is_finite());
        }
    }
}

#[test]
fn test_tsne_large_values() {
    // Test with large values (checks numerical stability)
    let data = Matrix::from_vec(4, 2, vec![1e6, 2e6, 1.1e6, 2.1e6, 5e6, 6e6, 5.1e6, 6.1e6])
        .expect("valid matrix dimensions");

    let mut tsne = TSNE::new(2).with_n_iter(50).with_random_state(42);
    let result = tsne
        .fit_transform(&data)
        .expect("fit_transform should succeed");
    assert_eq!(result.shape(), (4, 2));

    // Results should be finite
    for i in 0..4 {
        for j in 0..2 {
            assert!(result.get(i, j).is_finite());
        }
    }
}

#[test]
fn test_tsne_without_random_state() {
    let data = Matrix::from_vec(4, 2, vec![1.0, 2.0, 2.0, 3.0, 10.0, 11.0, 11.0, 12.0])
        .expect("valid matrix dimensions");

    // Without random state, should use time-based seed
    let mut tsne = TSNE::new(2).with_n_iter(10);
    let result = tsne
        .fit_transform(&data)
        .expect("fit_transform should succeed");
    assert_eq!(result.shape(), (4, 2));
}

// ========================================================================
// StandardScaler save/load edge cases
// ========================================================================

#[test]
fn test_standard_scaler_save_with_options() {
    use std::fs;
    use std::path::PathBuf;

    let data = Matrix::from_vec(3, 2, vec![1.0, 10.0, 2.0, 20.0, 3.0, 30.0])
        .expect("valid matrix dimensions");

    // Test with mean=false, std=true
    let mut scaler = StandardScaler::new().with_mean(false).with_std(true);
    scaler.fit(&data).expect("fit should succeed");

    let path = PathBuf::from("/tmp/test_standard_scaler_options.safetensors");
    scaler.save_safetensors(&path).expect("save should succeed");

    let loaded = StandardScaler::load_safetensors(&path).expect("load should succeed");

    // Verify transform behavior matches
    let test_data = Matrix::from_vec(1, 2, vec![2.0, 20.0]).expect("valid matrix dimensions");
    let orig_transformed = scaler
        .transform(&test_data)
        .expect("transform should succeed");
    let loaded_transformed = loaded
        .transform(&test_data)
        .expect("transform should succeed");

    assert!((orig_transformed.get(0, 0) - loaded_transformed.get(0, 0)).abs() < 1e-5);
    assert!((orig_transformed.get(0, 1) - loaded_transformed.get(0, 1)).abs() < 1e-5);

    let _ = fs::remove_file(&path);
}