anofox-forecast 0.15.9

Time series forecasting library
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

//! Signal carrier for changepoint detection.
//!
//! Wraps a flat `&[f64]` slice with `n × d` shape information so cost
//! functions and detectors can transparently support univariate and
//! multivariate signals.

use crate::error::{ForecastError, Result};

/// A multivariate signal of shape `(n, d)`, stored row-major as a flat
/// `&[f64]` slice. `n` is the number of samples, `d` is the dimension.
///
/// Univariate series can be wrapped via [`Signal::univariate`] or by
/// `Signal::from(values)`.
#[derive(Debug, Clone, Copy)]
pub struct Signal<'a> {
    data: &'a [f64],
    n: usize,
    d: usize,
}

impl<'a> Signal<'a> {
    /// Wrap a univariate sequence (`d = 1`).
    pub fn univariate(values: &'a [f64]) -> Self {
        Self {
            data: values,
            n: values.len(),
            d: 1,
        }
    }

    /// Wrap a multivariate signal stored row-major: `data[i*d + j]` is
    /// the j-th coordinate of the i-th sample.
    pub fn from_row_major(data: &'a [f64], n: usize, d: usize) -> Result<Self> {
        if d == 0 {
            return Err(ForecastError::InvalidParameter(
                "Signal: dimension d must be ≥ 1".into(),
            ));
        }
        if data.len() != n * d {
            return Err(ForecastError::DimensionMismatch {
                expected: n * d,
                got: data.len(),
            });
        }
        Ok(Self { data, n, d })
    }

    /// Number of samples.
    pub fn n(&self) -> usize {
        self.n
    }

    /// Dimension (number of features per sample).
    pub fn d(&self) -> usize {
        self.d
    }

    /// True iff `d == 1`.
    pub fn is_univariate(&self) -> bool {
        self.d == 1
    }

    /// Row `i` as a `&[f64]` of length `d`.
    ///
    /// # Panics
    /// Panics if `i >= n`.
    pub fn row(&self, i: usize) -> &[f64] {
        let start = i * self.d;
        let end = start + self.d;
        &self.data[start..end]
    }

    /// Full row-major slice.
    pub fn as_slice(&self) -> &[f64] {
        self.data
    }
}

impl<'a> From<&'a [f64]> for Signal<'a> {
    fn from(values: &'a [f64]) -> Self {
        Signal::univariate(values)
    }
}

impl<'a, const N: usize> From<&'a [f64; N]> for Signal<'a> {
    fn from(values: &'a [f64; N]) -> Self {
        Signal::univariate(values.as_slice())
    }
}

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

    #[test]
    fn univariate_from_slice_round_trip() {
        let values = [1.0, 2.0, 3.0, 4.0];
        let s: Signal = (&values[..]).into();
        assert_eq!(s.n(), 4);
        assert_eq!(s.d(), 1);
        assert!(s.is_univariate());
        assert_eq!(s.row(2), &[3.0]);
    }

    #[test]
    fn multivariate_row_major() {
        // n=3 samples, d=2 dims: [1,2, 3,4, 5,6]
        let data = vec![1.0, 2.0, 3.0, 4.0, 5.0, 6.0];
        let s = Signal::from_row_major(&data, 3, 2).unwrap();
        assert_eq!(s.n(), 3);
        assert_eq!(s.d(), 2);
        assert!(!s.is_univariate());
        assert_eq!(s.row(0), &[1.0, 2.0]);
        assert_eq!(s.row(1), &[3.0, 4.0]);
        assert_eq!(s.row(2), &[5.0, 6.0]);
    }

    #[test]
    fn dimension_mismatch_errors() {
        let data = vec![1.0, 2.0, 3.0];
        let err = Signal::from_row_major(&data, 2, 2).unwrap_err();
        assert!(matches!(err, ForecastError::DimensionMismatch { .. }));
    }

    #[test]
    fn zero_dimension_rejected() {
        let data = vec![1.0, 2.0];
        let err = Signal::from_row_major(&data, 2, 0).unwrap_err();
        assert!(matches!(err, ForecastError::InvalidParameter(_)));
    }
}