oxits 0.1.0

Time series classification and transformation library for 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
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
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388

use crate::core::traits::Transformer;

// --- StandardScaler ---

#[derive(Debug, Clone, Copy)]
pub struct StandardScalerConfig {
    pub with_mean: bool,
    pub with_std: bool,
}

impl StandardScalerConfig {
    pub fn new() -> Self {
        Self {
            with_mean: true,
            with_std: true,
        }
    }
}

impl Default for StandardScalerConfig {
    fn default() -> Self {
        Self::new()
    }
}

pub struct StandardScaler;

impl Transformer for StandardScaler {
    type Config = StandardScalerConfig;

    fn transform(config: &Self::Config, x: &[Vec<f64>]) -> Vec<Vec<f64>> {
        assert!(!x.is_empty(), "Input must have at least one sample");
        assert!(
            x.iter().all(|s| s.len() == x[0].len()),
            "All samples must have same length"
        );

        let with_mean = config.with_mean;
        let with_std = config.with_std;

        #[cfg(feature = "parallel")]
        {
            use rayon::prelude::*;
            return x
                .par_iter()
                .map(|sample| standard_scale_single(sample, with_mean, with_std))
                .collect();
        }

        #[cfg(not(feature = "parallel"))]
        x.iter()
            .map(|sample| standard_scale_single(sample, with_mean, with_std))
            .collect()
    }
}

fn standard_scale_single(x: &[f64], with_mean: bool, with_std: bool) -> Vec<f64> {
    let n = x.len() as f64;
    // Always compute actual mean for variance calculation
    let actual_mean = x.iter().sum::<f64>() / n;
    let center = if with_mean { actual_mean } else { 0.0 };
    let inv_std = if with_std {
        let variance = x
            .iter()
            .map(|&v| (v - actual_mean) * (v - actual_mean))
            .sum::<f64>()
            / n;
        let s = variance.sqrt();
        if s == 0.0 {
            1.0
        } else {
            1.0 / s
        }
    } else {
        1.0
    };
    x.iter().map(|&v| (v - center) * inv_std).collect()
}

// --- MinMaxScaler ---

#[derive(Debug, Clone, Copy)]
pub struct MinMaxScalerConfig {
    pub sample_range: (f64, f64),
}

impl MinMaxScalerConfig {
    pub fn new() -> Self {
        Self {
            sample_range: (0.0, 1.0),
        }
    }
}

impl Default for MinMaxScalerConfig {
    fn default() -> Self {
        Self::new()
    }
}

pub struct MinMaxScaler;

impl Transformer for MinMaxScaler {
    type Config = MinMaxScalerConfig;

    fn transform(config: &Self::Config, x: &[Vec<f64>]) -> Vec<Vec<f64>> {
        assert!(!x.is_empty(), "Input must have at least one sample");
        assert!(
            config.sample_range.0 < config.sample_range.1,
            "sample_range min must be less than max"
        );

        let (range_min, range_max) = config.sample_range;
        let scale_range = |sample: &Vec<f64>| {
            let (x_min, x_max) = sample
                .iter()
                .copied()
                .fold((f64::INFINITY, f64::NEG_INFINITY), |(mn, mx), v| {
                    (mn.min(v), mx.max(v))
                });
            let data_range = x_max - x_min;
            if data_range == 0.0 {
                vec![range_min; sample.len()]
            } else {
                let scale = (range_max - range_min) / data_range;
                sample
                    .iter()
                    .map(|&v| (v - x_min) * scale + range_min)
                    .collect()
            }
        };

        #[cfg(feature = "parallel")]
        {
            use rayon::prelude::*;
            return x.par_iter().map(scale_range).collect();
        }

        #[cfg(not(feature = "parallel"))]
        x.iter().map(scale_range).collect()
    }
}

// --- MaxAbsScaler ---

pub struct MaxAbsScaler;

impl Transformer for MaxAbsScaler {
    type Config = ();

    fn transform(_config: &Self::Config, x: &[Vec<f64>]) -> Vec<Vec<f64>> {
        assert!(!x.is_empty(), "Input must have at least one sample");

        x.iter()
            .map(|sample| {
                let max_abs = sample
                    .iter()
                    .copied()
                    .fold(0.0_f64, |acc, v| acc.max(v.abs()));
                let scale = if max_abs == 0.0 { 1.0 } else { max_abs };
                sample.iter().map(|&v| v / scale).collect()
            })
            .collect()
    }
}

// --- RobustScaler ---

#[derive(Debug, Clone, Copy)]
pub struct RobustScalerConfig {
    pub with_centering: bool,
    pub with_scaling: bool,
    pub quantile_range: (f64, f64),
}

impl RobustScalerConfig {
    pub fn new() -> Self {
        Self {
            with_centering: true,
            with_scaling: true,
            quantile_range: (25.0, 75.0),
        }
    }
}

impl Default for RobustScalerConfig {
    fn default() -> Self {
        Self::new()
    }
}

pub struct RobustScaler;

impl Transformer for RobustScaler {
    type Config = RobustScalerConfig;

    fn transform(config: &Self::Config, x: &[Vec<f64>]) -> Vec<Vec<f64>> {
        assert!(!x.is_empty(), "Input must have at least one sample");
        assert!(
            config.quantile_range.0 < config.quantile_range.1,
            "quantile_range lower must be less than upper"
        );
        assert!(
            config.quantile_range.0 >= 0.0 && config.quantile_range.1 <= 100.0,
            "quantile_range must be in [0, 100]"
        );

        x.iter()
            .map(|sample| {
                let center = if config.with_centering {
                    percentile(sample, 50.0)
                } else {
                    0.0
                };
                let scale = if config.with_scaling {
                    let q_lo = percentile(sample, config.quantile_range.0);
                    let q_hi = percentile(sample, config.quantile_range.1);
                    let iqr = q_hi - q_lo;
                    if iqr == 0.0 {
                        1.0
                    } else {
                        iqr
                    }
                } else {
                    1.0
                };
                sample.iter().map(|&v| (v - center) / scale).collect()
            })
            .collect()
    }
}

/// Compute the p-th percentile (0-100) of a slice using linear interpolation.
/// Matches numpy's default `interpolation='linear'`.
pub fn percentile(data: &[f64], p: f64) -> f64 {
    assert!(!data.is_empty(), "Cannot compute percentile of empty slice");
    assert!((0.0..=100.0).contains(&p), "Percentile must be in [0, 100]");

    let mut sorted: Vec<f64> = data.to_vec();
    sorted.sort_by(|a, b| a.partial_cmp(b).unwrap());

    if sorted.len() == 1 {
        return sorted[0];
    }

    let n = sorted.len();
    let idx = p / 100.0 * (n - 1) as f64;
    let lo = idx.floor() as usize;
    let hi = lo + 1;
    let frac = idx - lo as f64;

    if hi >= n {
        sorted[n - 1]
    } else {
        sorted[lo] + frac * (sorted[hi] - sorted[lo])
    }
}

#[cfg(test)]
mod tests {
    use super::*;

    fn assert_close(actual: &[f64], expected: &[f64], eps: f64) {
        assert_eq!(actual.len(), expected.len());
        for (i, (a, e)) in actual.iter().zip(expected.iter()).enumerate() {
            assert!(
                (a - e).abs() < eps,
                "[{i}]: actual={a}, expected={e}, eps={eps}"
            );
        }
    }

    #[test]
    fn test_standard_scaler_basic() {
        let config = StandardScalerConfig::new();
        let x = vec![vec![1.0, 2.0, 3.0, 4.0, 5.0]];
        let result = StandardScaler::transform(&config, &x);
        // mean=3.0, population std=sqrt(2.0)
        let std = 2.0_f64.sqrt();
        let expected: Vec<f64> = (1..=5).map(|v| (v as f64 - 3.0) / std).collect();
        assert_close(&result[0], &expected, 1e-10);
    }

    #[test]
    fn test_standard_scaler_no_mean() {
        let config = StandardScalerConfig {
            with_mean: false,
            with_std: true,
        };
        let x = vec![vec![1.0, 2.0, 3.0]];
        let result = StandardScaler::transform(&config, &x);
        // Variance is computed from actual mean (2.0): var = 2/3
        // But mean is NOT subtracted from output
        let var: f64 = 2.0 / 3.0;
        let std = var.sqrt();
        let expected: Vec<f64> = vec![1.0 / std, 2.0 / std, 3.0 / std];
        assert_close(&result[0], &expected, 1e-10);
    }

    #[test]
    fn test_standard_scaler_constant() {
        let config = StandardScalerConfig::new();
        let x = vec![vec![5.0, 5.0, 5.0]];
        let result = StandardScaler::transform(&config, &x);
        assert_close(&result[0], &[0.0, 0.0, 0.0], 1e-10);
    }

    #[test]
    fn test_minmax_scaler_basic() {
        let config = MinMaxScalerConfig::new();
        let x = vec![vec![1.0, 2.0, 3.0, 4.0, 5.0]];
        let result = MinMaxScaler::transform(&config, &x);
        assert_close(&result[0], &[0.0, 0.25, 0.5, 0.75, 1.0], 1e-10);
    }

    #[test]
    fn test_minmax_scaler_custom_range() {
        let config = MinMaxScalerConfig {
            sample_range: (-1.0, 1.0),
        };
        let x = vec![vec![0.0, 5.0, 10.0]];
        let result = MinMaxScaler::transform(&config, &x);
        assert_close(&result[0], &[-1.0, 0.0, 1.0], 1e-10);
    }

    #[test]
    fn test_minmax_scaler_constant() {
        let config = MinMaxScalerConfig::new();
        let x = vec![vec![3.0, 3.0, 3.0]];
        let result = MinMaxScaler::transform(&config, &x);
        // Constant row: fill with range_min (0.0)
        assert_close(&result[0], &[0.0, 0.0, 0.0], 1e-10);
    }

    #[test]
    fn test_maxabs_scaler_basic() {
        let x = vec![vec![-3.0, 1.0, 2.0]];
        let result = MaxAbsScaler::transform(&(), &x);
        assert_close(&result[0], &[-1.0, 1.0 / 3.0, 2.0 / 3.0], 1e-10);
    }

    #[test]
    fn test_maxabs_scaler_zero() {
        let x = vec![vec![0.0, 0.0, 0.0]];
        let result = MaxAbsScaler::transform(&(), &x);
        assert_close(&result[0], &[0.0, 0.0, 0.0], 1e-10);
    }

    #[test]
    fn test_robust_scaler_basic() {
        let config = RobustScalerConfig::new();
        let x = vec![vec![1.0, 2.0, 3.0, 4.0, 5.0]];
        let result = RobustScaler::transform(&config, &x);
        let median = 3.0;
        let q25 = percentile(&x[0], 25.0);
        let q75 = percentile(&x[0], 75.0);
        let iqr = q75 - q25;
        let expected: Vec<f64> = x[0].iter().map(|&v| (v - median) / iqr).collect();
        assert_close(&result[0], &expected, 1e-10);
    }

    #[test]
    fn test_percentile_basic() {
        let data = vec![1.0, 2.0, 3.0, 4.0, 5.0];
        assert!((percentile(&data, 0.0) - 1.0).abs() < 1e-10);
        assert!((percentile(&data, 50.0) - 3.0).abs() < 1e-10);
        assert!((percentile(&data, 100.0) - 5.0).abs() < 1e-10);
        assert!((percentile(&data, 25.0) - 2.0).abs() < 1e-10);
        assert!((percentile(&data, 75.0) - 4.0).abs() < 1e-10);
    }

    #[test]
    fn test_percentile_interpolation() {
        let data = vec![0.0, 10.0];
        assert!((percentile(&data, 50.0) - 5.0).abs() < 1e-10);
        assert!((percentile(&data, 25.0) - 2.5).abs() < 1e-10);
    }

    #[test]
    fn test_multiple_samples() {
        let config = StandardScalerConfig::new();
        let x = vec![vec![1.0, 2.0, 3.0], vec![10.0, 20.0, 30.0]];
        let result = StandardScaler::transform(&config, &x);
        assert_eq!(result.len(), 2);
        // Both should be scaled independently to same shape
        assert_close(&result[0], &result[1], 1e-10);
    }
}