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

use rand::rngs::StdRng;
use rand::{Rng, SeedableRng};
use rand_distr::Normal;
use std::collections::HashMap;

/// Learning Shapelets classifier.
///
/// Learns shapelets through gradient descent optimization:
/// 1. Initialize shapelets (random subsequences or K-means)
/// 2. Compute shapelet distances as features
/// 3. Optimize shapelets + logistic regression weights via SGD
/// 4. Classify using the learned model

#[derive(Debug, Clone)]
pub struct LearningShapeletsConfig {
    pub n_shapelets_per_size: usize,
    pub shapelet_sizes: Vec<usize>,
    pub learning_rate: f64,
    pub n_epochs: usize,
    pub regularization: f64,
    pub random_seed: Option<u64>,
}

impl LearningShapeletsConfig {
    pub fn new() -> Self {
        Self {
            n_shapelets_per_size: 3,
            shapelet_sizes: vec![3, 5],
            learning_rate: 0.01,
            n_epochs: 100,
            regularization: 0.01,
            random_seed: None,
        }
    }
}

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

#[derive(Debug, Clone)]
pub struct LearningShapeletsFitted {
    pub shapelets: Vec<Vec<f64>>,
    pub weights: Vec<Vec<f64>>, // (n_classes, n_shapelets)
    pub biases: Vec<f64>,       // (n_classes,)
    pub classes: Vec<String>,
}

pub struct LearningShapelets;

impl LearningShapelets {
    /// Fit the Learning Shapelets model via gradient descent.
    pub fn fit(
        config: &LearningShapeletsConfig,
        x: &[Vec<f64>],
        y: &[String],
    ) -> LearningShapeletsFitted {
        assert!(!x.is_empty(), "Input must have at least one sample");
        assert_eq!(x.len(), y.len(), "X and y must have same length");

        let n_timestamps = x[0].len();
        let mut rng = match config.random_seed {
            Some(seed) => StdRng::seed_from_u64(seed),
            None => StdRng::from_entropy(),
        };

        // Get unique classes
        let mut classes: Vec<String> = y.to_vec();
        classes.sort();
        classes.dedup();
        let n_classes = classes.len();
        let class_idx: HashMap<&str, usize> = classes
            .iter()
            .enumerate()
            .map(|(i, c)| (c.as_str(), i))
            .collect();

        // Initialize shapelets from random subsequences
        let mut shapelets: Vec<Vec<f64>> = Vec::new();
        for &size in &config.shapelet_sizes {
            let size = size.min(n_timestamps);
            for _ in 0..config.n_shapelets_per_size {
                let sample_idx = rng.gen_range(0..x.len());
                let start = rng.gen_range(0..=n_timestamps - size);
                let shapelet = x[sample_idx][start..start + size].to_vec();
                shapelets.push(shapelet);
            }
        }
        let n_shapelets = shapelets.len();

        // Initialize weights with small random values
        let normal = Normal::new(0.0, 0.01).unwrap();
        let mut weights: Vec<Vec<f64>> = (0..n_classes)
            .map(|_| (0..n_shapelets).map(|_| rng.sample(normal)).collect())
            .collect();
        let mut biases = vec![0.0; n_classes];

        // Encode labels as one-hot
        let targets: Vec<Vec<f64>> = y
            .iter()
            .map(|label| {
                let mut one_hot = vec![0.0; n_classes];
                one_hot[class_idx[label.as_str()]] = 1.0;
                one_hot
            })
            .collect();

        // SGD optimization
        let lr = config.learning_rate;
        let reg = config.regularization;

        for _epoch in 0..config.n_epochs {
            // Shuffle indices
            let mut indices: Vec<usize> = (0..x.len()).collect();
            for i in (1..indices.len()).rev() {
                let j = rng.gen_range(0..=i);
                indices.swap(i, j);
            }

            for &idx in &indices {
                // Forward pass: compute shapelet distances
                let distances: Vec<f64> = shapelets
                    .iter()
                    .map(|s| min_soft_distance(&x[idx], s))
                    .collect();

                // Compute logits: W * distances + b
                let logits: Vec<f64> = (0..n_classes)
                    .map(|c| {
                        weights[c]
                            .iter()
                            .zip(distances.iter())
                            .map(|(&w, &d)| w * d)
                            .sum::<f64>()
                            + biases[c]
                    })
                    .collect();

                // Softmax
                let max_logit = logits.iter().copied().fold(f64::NEG_INFINITY, f64::max);
                let exp: Vec<f64> = logits.iter().map(|&l| (l - max_logit).exp()).collect();
                let sum_exp: f64 = exp.iter().sum();
                let probs: Vec<f64> = exp.iter().map(|&e| e / sum_exp).collect();

                // Gradient: dL/dlogit = probs - targets
                let d_logits: Vec<f64> = probs
                    .iter()
                    .zip(targets[idx].iter())
                    .map(|(&p, &t)| p - t)
                    .collect();

                // Update weights and biases
                for c in 0..n_classes {
                    for s in 0..n_shapelets {
                        weights[c][s] -= lr * (d_logits[c] * distances[s] + reg * weights[c][s]);
                    }
                    biases[c] -= lr * d_logits[c];
                }

                // Update shapelets via chain rule
                // dL/dshapelet_s = sum_c(dL/dlogit_c * w_cs) * d(distance_s)/d(shapelet_s)
                for s in 0..n_shapelets {
                    let d_dist: f64 = (0..n_classes).map(|c| d_logits[c] * weights[c][s]).sum();

                    // Gradient of soft-min distance w.r.t. shapelet
                    let grad = shapelet_gradient(&x[idx], &shapelets[s]);
                    for (k, g) in grad.iter().enumerate() {
                        shapelets[s][k] -= lr * d_dist * g;
                    }
                }
            }
        }

        LearningShapeletsFitted {
            shapelets,
            weights,
            biases,
            classes,
        }
    }

    /// Predict class labels.
    pub fn predict(fitted: &LearningShapeletsFitted, x: &[Vec<f64>]) -> Vec<String> {
        let n_classes = fitted.classes.len();

        x.iter()
            .map(|sample| {
                let distances: Vec<f64> = fitted
                    .shapelets
                    .iter()
                    .map(|s| min_soft_distance(sample, s))
                    .collect();

                // Compute logits
                let logits: Vec<f64> = (0..n_classes)
                    .map(|c| {
                        fitted.weights[c]
                            .iter()
                            .zip(distances.iter())
                            .map(|(&w, &d)| w * d)
                            .sum::<f64>()
                            + fitted.biases[c]
                    })
                    .collect();

                // Argmax
                let (best_class, _) = logits
                    .iter()
                    .enumerate()
                    .max_by(|a, b| a.1.partial_cmp(b.1).unwrap())
                    .unwrap();

                fitted.classes[best_class].clone()
            })
            .collect()
    }

    /// Compute classification accuracy.
    pub fn score(fitted: &LearningShapeletsFitted, x: &[Vec<f64>], y: &[String]) -> f64 {
        let predictions = Self::predict(fitted, x);
        let correct = predictions
            .iter()
            .zip(y.iter())
            .filter(|(p, t)| p == t)
            .count();
        correct as f64 / y.len() as f64
    }
}

/// Compute soft-minimum distance between time series and shapelet.
/// Uses the log-sum-exp trick for smooth minimum.
fn min_soft_distance(ts: &[f64], shapelet: &[f64]) -> f64 {
    let n = ts.len();
    let m = shapelet.len();
    if m > n {
        return f64::INFINITY;
    }

    let alpha = -10.0; // smoothing parameter (more negative = closer to true min)

    let mut distances = Vec::with_capacity(n - m + 1);
    for i in 0..=n - m {
        let d: f64 = ts[i..i + m]
            .iter()
            .zip(shapelet.iter())
            .map(|(&a, &b)| (a - b).powi(2))
            .sum::<f64>()
            / m as f64;
        distances.push(d);
    }

    // Soft-min using log-sum-exp
    let max_alpha_d = distances
        .iter()
        .map(|&d| alpha * d)
        .fold(f64::NEG_INFINITY, f64::max);
    let sum_exp: f64 = distances
        .iter()
        .map(|&d| (alpha * d - max_alpha_d).exp())
        .sum();
    (max_alpha_d + sum_exp.ln()) / alpha
}

/// Compute gradient of soft-min distance w.r.t. shapelet values.
fn shapelet_gradient(ts: &[f64], shapelet: &[f64]) -> Vec<f64> {
    let n = ts.len();
    let m = shapelet.len();
    if m > n {
        return vec![0.0; m];
    }

    let alpha = -10.0;

    // Compute per-position distances
    let mut distances = Vec::with_capacity(n - m + 1);
    for i in 0..=n - m {
        let d: f64 = ts[i..i + m]
            .iter()
            .zip(shapelet.iter())
            .map(|(&a, &b)| (a - b).powi(2))
            .sum::<f64>()
            / m as f64;
        distances.push(d);
    }

    // Softmax weights (attention over positions)
    let max_alpha_d = distances
        .iter()
        .map(|&d| alpha * d)
        .fold(f64::NEG_INFINITY, f64::max);
    let exp_weights: Vec<f64> = distances
        .iter()
        .map(|&d| (alpha * d - max_alpha_d).exp())
        .collect();
    let sum_exp: f64 = exp_weights.iter().sum();
    let softmax_weights: Vec<f64> = exp_weights.iter().map(|&e| e / sum_exp).collect();

    // Gradient: weighted average of per-position gradients
    let mut grad = vec![0.0; m];
    for (i, &w) in softmax_weights.iter().enumerate() {
        for (k, g) in grad.iter_mut().enumerate() {
            // d/d(shapelet_k) of ||ts[i..] - shapelet||^2 / m
            // = -2 * (ts[i+k] - shapelet_k) / m
            *g += w * (-2.0 * (ts[i + k] - shapelet[k]) / m as f64);
        }
    }

    grad
}

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

    #[test]
    fn test_learning_shapelets_basic() {
        let config = LearningShapeletsConfig {
            n_shapelets_per_size: 2,
            shapelet_sizes: vec![3],
            learning_rate: 0.1,
            n_epochs: 50,
            regularization: 0.001,
            random_seed: Some(42),
        };
        let x = vec![
            vec![0.0, 1.0, 2.0, 3.0, 4.0, 5.0],
            vec![0.0, 1.0, 2.0, 3.0, 4.0, 6.0],
            vec![5.0, 4.0, 3.0, 2.0, 1.0, 0.0],
            vec![6.0, 4.0, 3.0, 2.0, 1.0, 0.0],
        ];
        let y = vec![
            "A".to_string(),
            "A".to_string(),
            "B".to_string(),
            "B".to_string(),
        ];
        let fitted = LearningShapelets::fit(&config, &x, &y);
        let predictions = LearningShapelets::predict(&fitted, &x);
        assert_eq!(predictions.len(), 4);
    }

    #[test]
    fn test_learning_shapelets_converges() {
        let config = LearningShapeletsConfig {
            n_shapelets_per_size: 3,
            shapelet_sizes: vec![3, 4],
            learning_rate: 0.1,
            n_epochs: 200,
            regularization: 0.001,
            random_seed: Some(42),
        };
        // Clearly separable data
        let x = vec![
            vec![0.0, 1.0, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0],
            vec![0.0, 0.5, 1.5, 2.5, 3.5, 4.5, 5.5, 7.5],
            vec![7.0, 6.0, 5.0, 4.0, 3.0, 2.0, 1.0, 0.0],
            vec![7.5, 5.5, 4.5, 3.5, 2.5, 1.5, 0.5, 0.0],
        ];
        let y = vec![
            "up".to_string(),
            "up".to_string(),
            "down".to_string(),
            "down".to_string(),
        ];
        let fitted = LearningShapelets::fit(&config, &x, &y);
        let score = LearningShapelets::score(&fitted, &x, &y);
        // Should be able to fit training data well
        assert!(score >= 0.5, "Score too low: {score}");
    }

    #[test]
    fn test_soft_min_distance() {
        let ts = vec![0.0, 0.0, 0.0, 1.0, 1.0, 1.0];
        let shapelet = vec![0.0, 0.0, 0.0];
        let d = min_soft_distance(&ts, &shapelet);
        // Soft-min can be slightly negative due to smoothing, but close to 0
        assert!(
            d < 0.1,
            "Distance should be near 0 for exact match, got {d}"
        );
        assert!(d > -0.1, "Distance too negative: {d}");
    }
}