aprender-core 0.30.0

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
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
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303

pub(crate) use super::*;

#[test]
fn test_cov_positive_relationship() {
    let x = Vector::from_slice(&[1.0, 2.0, 3.0, 4.0, 5.0]);
    let y = Vector::from_slice(&[2.0, 4.0, 6.0, 8.0, 10.0]);

    let covariance = cov(&x, &y).expect("Should compute covariance");
    assert!(
        covariance > 0.0,
        "Positive relationship should have positive covariance"
    );
}

#[test]
fn test_cov_negative_relationship() {
    let x = Vector::from_slice(&[1.0, 2.0, 3.0, 4.0, 5.0]);
    let y = Vector::from_slice(&[10.0, 8.0, 6.0, 4.0, 2.0]);

    let covariance = cov(&x, &y).expect("Should compute covariance");
    assert!(
        covariance < 0.0,
        "Negative relationship should have negative covariance"
    );
}

#[test]
fn test_cov_dimension_mismatch() {
    let x = Vector::from_slice(&[1.0, 2.0, 3.0]);
    let y = Vector::from_slice(&[1.0, 2.0]);

    let result = cov(&x, &y);
    assert!(result.is_err());
    let err = result.expect_err("Should be dimension mismatch");
    assert!(matches!(err, AprenderError::DimensionMismatch { .. }));
}

#[test]
fn test_cov_empty() {
    let x = Vector::from_slice(&[]);
    let y = Vector::from_slice(&[]);

    let result = cov(&x, &y);
    assert!(result.is_err());
}

#[test]
fn test_cov_matrix_simple() {
    // 3 samples, 2 features
    let data = Matrix::from_vec(3, 2, vec![1.0, 2.0, 2.0, 4.0, 3.0, 6.0]).expect("Valid matrix");

    let cov_mat = cov_matrix(&data).expect("Should compute covariance matrix");

    assert_eq!(cov_mat.n_rows(), 2);
    assert_eq!(cov_mat.n_cols(), 2);

    // Covariance matrix should be symmetric
    assert!((cov_mat.get(0, 1) - cov_mat.get(1, 0)).abs() < 1e-6);

    // Diagonal should be positive (variances)
    assert!(cov_mat.get(0, 0) > 0.0);
    assert!(cov_mat.get(1, 1) > 0.0);
}

#[test]
fn test_corr_perfect_positive() {
    let x = Vector::from_slice(&[1.0, 2.0, 3.0, 4.0]);
    let y = Vector::from_slice(&[2.0, 4.0, 6.0, 8.0]);

    let correlation = corr(&x, &y).expect("Should compute correlation");
    assert!(
        (correlation - 1.0).abs() < 1e-6,
        "Perfect positive correlation should be 1.0"
    );
}

#[test]
fn test_corr_perfect_negative() {
    let x = Vector::from_slice(&[1.0, 2.0, 3.0, 4.0]);
    let y = Vector::from_slice(&[8.0, 6.0, 4.0, 2.0]);

    let correlation = corr(&x, &y).expect("Should compute correlation");
    assert!(
        (correlation + 1.0).abs() < 1e-6,
        "Perfect negative correlation should be -1.0"
    );
}

#[test]
fn test_corr_no_relationship() {
    // Orthogonal vectors (mean-centered)
    let x = Vector::from_slice(&[-1.0, 0.0, 1.0]);
    let y = Vector::from_slice(&[1.0, -2.0, 1.0]);

    let correlation = corr(&x, &y).expect("Should compute correlation");
    assert!(
        correlation.abs() < 0.1,
        "Orthogonal vectors should have near-zero correlation, got {correlation}"
    );
}

#[test]
fn test_corr_zero_variance() {
    let x = Vector::from_slice(&[1.0, 1.0, 1.0]);
    let y = Vector::from_slice(&[2.0, 4.0, 6.0]);

    let result = corr(&x, &y);
    assert!(result.is_err(), "Should error when variance is zero");
}

#[test]
fn test_corr_matrix_simple() {
    // 4 samples, 2 features
    let data =
        Matrix::from_vec(4, 2, vec![1.0, 2.0, 2.0, 4.0, 3.0, 6.0, 4.0, 8.0]).expect("Valid matrix");

    let corr_mat = corr_matrix(&data).expect("Should compute correlation matrix");

    assert_eq!(corr_mat.n_rows(), 2);
    assert_eq!(corr_mat.n_cols(), 2);

    // Diagonal should be 1.0
    assert!((corr_mat.get(0, 0) - 1.0).abs() < 1e-6);
    assert!((corr_mat.get(1, 1) - 1.0).abs() < 1e-6);

    // Should be symmetric
    assert!((corr_mat.get(0, 1) - corr_mat.get(1, 0)).abs() < 1e-6);

    // Perfect correlation (y = 2*x)
    assert!((corr_mat.get(0, 1) - 1.0).abs() < 1e-6);
}

#[test]
fn test_corr_matrix_independent() {
    // 3 samples, 2 independent features
    let data = Matrix::from_vec(3, 2, vec![1.0, 1.0, 2.0, 1.0, 3.0, 1.0]).expect("Valid matrix");

    let result = corr_matrix(&data);
    // Second feature has zero variance
    assert!(result.is_err());
}

#[test]
fn test_corr_matrix_three_features() {
    // 3 samples, 3 features
    let data = Matrix::from_vec(3, 3, vec![1.0, 2.0, 3.0, 2.0, 3.0, 4.0, 3.0, 4.0, 5.0])
        .expect("Valid matrix");

    let corr_mat = corr_matrix(&data).expect("Should compute correlation matrix");

    assert_eq!(corr_mat.n_rows(), 3);
    assert_eq!(corr_mat.n_cols(), 3);

    // All diagonals should be 1.0
    for i in 0..3 {
        assert!((corr_mat.get(i, i) - 1.0).abs() < 1e-6);
    }

    // All features perfectly correlated (linear relationship)
    for i in 0..3 {
        for j in 0..3 {
            assert!((corr_mat.get(i, j) - 1.0).abs() < 1e-6);
        }
    }
}

// =========================================================================
// Additional coverage: error paths and edge cases
// =========================================================================

#[test]
fn test_corr_dimension_mismatch() {
    let x = Vector::from_slice(&[1.0, 2.0, 3.0]);
    let y = Vector::from_slice(&[1.0, 2.0]);
    let result = corr(&x, &y);
    assert!(result.is_err());
    let err = result.expect_err("Should be dimension mismatch");
    assert!(matches!(err, AprenderError::DimensionMismatch { .. }));
}

#[test]
fn test_corr_empty() {
    let x = Vector::from_slice(&[]);
    let y = Vector::from_slice(&[]);
    let result = corr(&x, &y);
    assert!(result.is_err());
}

#[test]
fn test_corr_both_zero_variance() {
    let x = Vector::from_slice(&[5.0, 5.0, 5.0]);
    let y = Vector::from_slice(&[3.0, 3.0, 3.0]);
    let result = corr(&x, &y);
    assert!(result.is_err(), "Should error when both have zero variance");
}

#[test]
fn test_corr_y_zero_variance() {
    let x = Vector::from_slice(&[1.0, 2.0, 3.0]);
    let y = Vector::from_slice(&[5.0, 5.0, 5.0]);
    let result = corr(&x, &y);
    assert!(result.is_err(), "Should error when y has zero variance");
}

#[test]
fn test_cov_matrix_empty_rows() {
    // 0 rows, 2 cols -> empty data
    let data = Matrix::from_vec(0, 0, vec![]);
    assert!(data.is_err() || cov_matrix(&data.expect("empty matrix")).is_err());
}

#[test]
fn test_corr_matrix_empty() {
    // Attempt to create an empty-like matrix. Since from_vec may reject 0x0,
    // test with a valid but single-row matrix where we cannot compute variance
    // meaningfully. Let's trigger the n==0||p==0 path differently.
    // Instead, just test with a constant-feature matrix for the zero-variance path.
    let data = Matrix::from_vec(3, 2, vec![1.0, 5.0, 2.0, 5.0, 3.0, 5.0]).expect("Valid matrix");
    let result = corr_matrix(&data);
    // Feature 1 (column index 1) has zero variance
    assert!(result.is_err());
}

#[test]
fn test_cov_single_element() {
    let x = Vector::from_slice(&[5.0]);
    let y = Vector::from_slice(&[10.0]);
    let result = cov(&x, &y).expect("Should compute covariance for single element");
    // Cov of single element is 0 (no deviation)
    assert!((result - 0.0).abs() < 1e-10);
}

#[test]
fn test_cov_zero_covariance() {
    // x and y are uncorrelated (mean-centered, orthogonal)
    let x = Vector::from_slice(&[-1.0, 0.0, 1.0]);
    let y = Vector::from_slice(&[1.0, -2.0, 1.0]);
    let result = cov(&x, &y).expect("Should compute covariance");
    assert!(
        result.abs() < 1e-6,
        "Covariance of orthogonal vectors should be ~0"
    );
}

#[test]
fn test_cov_matrix_single_feature() {
    // 3 samples, 1 feature => 1x1 covariance matrix
    let data = Matrix::from_vec(3, 1, vec![1.0, 2.0, 3.0]).expect("Valid matrix");
    let cov_mat = cov_matrix(&data).expect("Should compute 1x1 covariance matrix");
    assert_eq!(cov_mat.n_rows(), 1);
    assert_eq!(cov_mat.n_cols(), 1);
    // Variance of [1,2,3] = 2/3
    assert!((cov_mat.get(0, 0) - 2.0 / 3.0).abs() < 1e-5);
}

#[test]
fn test_cov_matrix_single_sample() {
    // 1 sample, 2 features => all zeros (no variance with 1 sample)
    let data = Matrix::from_vec(1, 2, vec![5.0, 10.0]).expect("Valid matrix");
    let cov_mat = cov_matrix(&data).expect("Should compute covariance matrix");
    assert_eq!(cov_mat.n_rows(), 2);
    assert_eq!(cov_mat.n_cols(), 2);
    assert!((cov_mat.get(0, 0) - 0.0).abs() < 1e-10);
    assert!((cov_mat.get(1, 1) - 0.0).abs() < 1e-10);
}

#[test]
fn test_corr_matrix_negative_correlation() {
    // Feature 1 increases, feature 2 decreases
    let data =
        Matrix::from_vec(4, 2, vec![1.0, 8.0, 2.0, 6.0, 3.0, 4.0, 4.0, 2.0]).expect("Valid matrix");
    let corr_mat = corr_matrix(&data).expect("Should compute correlation matrix");
    // Off-diagonal should be negative (inverse relationship)
    assert!(
        corr_mat.get(0, 1) < 0.0,
        "Should have negative correlation, got {}",
        corr_mat.get(0, 1)
    );
    // Diagonal should be 1.0
    assert!((corr_mat.get(0, 0) - 1.0).abs() < 1e-6);
    assert!((corr_mat.get(1, 1) - 1.0).abs() < 1e-6);
}

#[test]
fn test_corr_single_pair() {
    // Minimum valid correlation: 2 values each
    let x = Vector::from_slice(&[0.0, 1.0]);
    let y = Vector::from_slice(&[0.0, 1.0]);
    let result = corr(&x, &y).expect("Should compute correlation for 2 elements");
    assert!((result - 1.0).abs() < 1e-6, "Perfect positive correlation");
}

#[test]
fn test_cov_identical_vectors() {
    // Cov(X, X) = Var(X)
    let x = Vector::from_slice(&[1.0, 2.0, 3.0, 4.0, 5.0]);
    let cov_xx = cov(&x, &x).expect("Should compute self-covariance");
    // Population variance of [1,2,3,4,5] = 2.0
    assert!(
        (cov_xx - 2.0).abs() < 1e-5,
        "Cov(X,X) should equal Var(X), got {cov_xx}"
    );
}