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

use crate::approximation::dft::{anova_selection, Dft, DftConfig};
use crate::approximation::mcb::{Mcb, McbConfig, McbFitted};
use crate::core::config::BinStrategy;
use crate::core::traits::{FittableTransformer, Transformer};

/// Symbolic Fourier Approximation (SFA).
///
/// Pipeline: DFT (with optional ANOVA feature selection) → MCB.
/// This is a stateful transformer: fit() learns the DFT coefficient selection
/// and MCB bin edges.

#[derive(Debug, Clone)]
pub struct SfaConfig {
    pub n_coefs: Option<usize>,
    pub n_bins: usize,
    pub strategy: BinStrategy,
    pub drop_sum: bool,
    pub anova: bool,
    pub norm_mean: bool,
    pub norm_std: bool,
}

impl SfaConfig {
    pub fn new() -> Self {
        Self {
            n_coefs: None,
            n_bins: 4,
            strategy: BinStrategy::Quantile,
            drop_sum: false,
            anova: false,
            norm_mean: false,
            norm_std: false,
        }
    }
}

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

#[derive(Debug, Clone)]
pub struct SfaFitted {
    pub dft_config: DftConfig,
    pub support: Vec<usize>,
    pub mcb_fitted: McbFitted,
}

pub struct Sfa;

impl FittableTransformer for Sfa {
    type Config = SfaConfig;
    type Fitted = SfaFitted;

    fn fit(config: &Self::Config, x: &[Vec<f64>], y: Option<&[String]>) -> Self::Fitted {
        assert!(!x.is_empty(), "Input must have at least one sample");
        assert!(
            config.n_bins >= 2 && config.n_bins <= 26,
            "n_bins must be in [2, 26]"
        );

        let dft_config = DftConfig {
            n_coefs: None, // compute all first, then select
            norm_mean: config.norm_mean,
            norm_std: config.norm_std,
            drop_sum: config.drop_sum,
        };

        // Step 1: Compute full DFT
        let all_coefs = Dft::transform(&dft_config, x);
        let n_available = all_coefs[0].len();
        let n_coefs = config.n_coefs.unwrap_or(n_available).min(n_available);

        // Step 2: Feature selection
        let support = if config.anova {
            let y = y.expect("ANOVA selection requires labels (y)");
            anova_selection(x, y, n_coefs, &dft_config)
        } else {
            (0..n_coefs).collect()
        };

        // Step 3: Select coefficients
        let selected: Vec<Vec<f64>> = all_coefs
            .iter()
            .map(|coefs| support.iter().map(|&i| coefs[i]).collect())
            .collect();

        // Step 4: Fit MCB on selected coefficients
        let mcb_config = McbConfig {
            n_bins: config.n_bins,
            strategy: config.strategy,
        };
        let mcb_fitted = Mcb::fit(&mcb_config, &selected, None);

        SfaFitted {
            dft_config,
            support,
            mcb_fitted,
        }
    }

    fn transform(fitted: &Self::Fitted, x: &[Vec<f64>]) -> Vec<Vec<f64>> {
        // Step 1: Compute DFT
        let all_coefs = Dft::transform(&fitted.dft_config, x);

        // Step 2: Select coefficients
        let selected: Vec<Vec<f64>> = all_coefs
            .iter()
            .map(|coefs| fitted.support.iter().map(|&i| coefs[i]).collect())
            .collect();

        // Step 3: Discretize with fitted MCB
        Mcb::transform(&fitted.mcb_fitted, &selected)
    }
}

/// Fit SFA from pre-computed DFT coefficients (skips redundant DFT).
/// Used by BOSS fused fit_transform to avoid computing DFT twice.
pub(crate) fn fit_from_coefs(
    config: &SfaConfig,
    all_coefs: &[Vec<f64>],
    y: Option<&[String]>,
    dft_config: &DftConfig,
) -> SfaFitted {
    let n_available = all_coefs[0].len();
    let n_coefs = config.n_coefs.unwrap_or(n_available).min(n_available);

    // Feature selection
    let support: Vec<usize> = if config.anova {
        let y = y.expect("ANOVA selection requires labels (y)");
        // For ANOVA with pre-computed coefs, we need the original x — but BOSS
        // doesn't use ANOVA, so this path isn't hit in the fused case.
        // Fall back to first n_coefs indices.
        let _ = y;
        (0..n_coefs).collect()
    } else {
        (0..n_coefs).collect()
    };

    // Select coefficients
    let selected: Vec<Vec<f64>> = all_coefs
        .iter()
        .map(|coefs| support.iter().map(|&i| coefs[i]).collect())
        .collect();

    // Fit MCB on selected coefficients
    let mcb_config = McbConfig {
        n_bins: config.n_bins,
        strategy: config.strategy,
    };
    let mcb_fitted = Mcb::fit(&mcb_config, &selected, None);

    SfaFitted {
        dft_config: dft_config.clone(),
        support,
        mcb_fitted,
    }
}

/// Transform pre-computed DFT coefficients to symbolic output (skips redundant DFT).
/// Used by BOSS fused fit_transform.
pub(crate) fn transform_from_coefs(fitted: &SfaFitted, all_coefs: &[Vec<f64>]) -> Vec<Vec<u8>> {
    // Select coefficients
    let selected: Vec<Vec<f64>> = all_coefs
        .iter()
        .map(|coefs| fitted.support.iter().map(|&i| coefs[i]).collect())
        .collect();

    // Discretize with fitted MCB and convert to u8
    let float_result = Mcb::transform(&fitted.mcb_fitted, &selected);
    float_result
        .into_iter()
        .map(|row| row.into_iter().map(|v| v as u8).collect())
        .collect()
}

/// Transform to symbolic u8 output.
pub fn sfa_transform_symbolic(fitted: &SfaFitted, x: &[Vec<f64>]) -> Vec<Vec<u8>> {
    let float_result = Sfa::transform(fitted, x);
    float_result
        .into_iter()
        .map(|row| row.into_iter().map(|v| v as u8).collect())
        .collect()
}

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

    #[test]
    fn test_sfa_basic() {
        let config = SfaConfig {
            n_coefs: Some(4),
            n_bins: 4,
            ..SfaConfig::new()
        };
        let x = vec![
            vec![0.0, 1.0, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0],
            vec![7.0, 6.0, 5.0, 4.0, 3.0, 2.0, 1.0, 0.0],
            vec![0.0, 2.0, 4.0, 6.0, 4.0, 2.0, 0.0, -2.0],
            vec![1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0],
        ];
        let fitted = Sfa::fit(&config, &x, None);
        let result = Sfa::transform(&fitted, &x);
        assert_eq!(result.len(), 4);
        assert_eq!(result[0].len(), 4);
    }

    #[test]
    fn test_sfa_fit_transform_consistency() {
        let config = SfaConfig {
            n_coefs: Some(3),
            n_bins: 3,
            ..SfaConfig::new()
        };
        let x = vec![
            vec![1.0, 2.0, 3.0, 4.0],
            vec![4.0, 3.0, 2.0, 1.0],
            vec![2.0, 4.0, 2.0, 4.0],
        ];
        let ft = Sfa::fit_transform(&config, &x, None);
        let fitted = Sfa::fit(&config, &x, None);
        let t = Sfa::transform(&fitted, &x);
        assert_eq!(ft, t);
    }

    #[test]
    fn test_sfa_anova() {
        let config = SfaConfig {
            n_coefs: Some(2),
            n_bins: 3,
            anova: true,
            ..SfaConfig::new()
        };
        let x = vec![
            vec![1.0, 2.0, 3.0, 4.0],
            vec![4.0, 3.0, 2.0, 1.0],
            vec![1.0, 3.0, 3.0, 4.0],
            vec![4.0, 2.0, 2.0, 1.0],
        ];
        let y = vec![
            "A".to_string(),
            "B".to_string(),
            "A".to_string(),
            "B".to_string(),
        ];
        let fitted = Sfa::fit(&config, &x, Some(&y));
        assert_eq!(fitted.support.len(), 2);
        let result = Sfa::transform(&fitted, &x);
        assert_eq!(result[0].len(), 2);
    }

    #[test]
    fn test_sfa_symbolic() {
        let config = SfaConfig {
            n_coefs: Some(3),
            n_bins: 4,
            ..SfaConfig::new()
        };
        let x = vec![
            vec![1.0, 2.0, 3.0, 4.0, 5.0, 6.0],
            vec![6.0, 5.0, 4.0, 3.0, 2.0, 1.0],
        ];
        let fitted = Sfa::fit(&config, &x, None);
        let symbolic = sfa_transform_symbolic(&fitted, &x);
        assert_eq!(symbolic.len(), 2);
        for row in &symbolic {
            assert_eq!(row.len(), 3);
            for &v in row {
                assert!(v < 4);
            }
        }
    }
}