clump 0.5.6

Dense clustering primitives (k-means, DBSCAN, HDBSCAN, EVoC, COP-Kmeans, DenStream, correlation clustering)
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

//! Adapters for converting external data to clump's [`DataRef`](super::flat::DataRef) input.

/// Convert an ndarray `Array2<f32>` to clump's input format.
///
/// Each row becomes one data point.
///
/// ```ignore
/// use ndarray::Array2;
/// use clump::cluster::adapt::array2_to_vecs;
///
/// let arr = Array2::from_shape_vec((3, 2), vec![1.0, 2.0, 3.0, 4.0, 5.0, 6.0]).unwrap();
/// let vecs = array2_to_vecs(&arr);
/// assert_eq!(vecs.len(), 3);
/// assert_eq!(vecs[0], vec![1.0, 2.0]);
/// ```
#[cfg(feature = "ndarray")]
pub fn array2_to_vecs(arr: &ndarray::Array2<f32>) -> Vec<Vec<f32>> {
    arr.rows().into_iter().map(|row| row.to_vec()).collect()
}

/// Convert an ndarray `ArrayView2<f32>` to clump's input format.
#[cfg(feature = "ndarray")]
pub fn array_view_to_vecs(arr: ndarray::ArrayView2<f32>) -> Vec<Vec<f32>> {
    arr.rows().into_iter().map(|row| row.to_vec()).collect()
}

/// Convert a flat slice + dimensions to clump's input format.
///
/// ```
/// use clump::cluster::adapt::flat_to_vecs;
///
/// let flat = vec![1.0f32, 2.0, 3.0, 4.0, 5.0, 6.0];
/// let vecs = flat_to_vecs(&flat, 3, 2);
/// assert_eq!(vecs.len(), 3);
/// assert_eq!(vecs[1], vec![3.0, 4.0]);
/// ```
pub fn flat_to_vecs(flat: &[f32], n: usize, d: usize) -> Vec<Vec<f32>> {
    assert_eq!(flat.len(), n * d, "flat.len() must equal n * d");
    (0..n).map(|i| flat[i * d..(i + 1) * d].to_vec()).collect()
}

/// Convert clump labels to an ndarray `Array1<usize>`.
#[cfg(feature = "ndarray")]
pub fn labels_to_array1(labels: &[usize]) -> ndarray::Array1<usize> {
    ndarray::Array1::from(labels.to_vec())
}

/// Create a [`FlatRef`](super::flat::FlatRef) from a flat slice + dimensions.
pub fn flat_to_ref(flat: &[f32], n: usize, d: usize) -> super::flat::FlatRef<'_> {
    super::flat::FlatRef::new(flat, n, d)
}

#[cfg(feature = "ndarray")]
impl super::flat::DataRef for ndarray::ArrayView2<'_, f32> {
    fn n(&self) -> usize {
        self.nrows()
    }
    fn d(&self) -> usize {
        self.ncols()
    }
    fn row(&self, i: usize) -> &[f32] {
        let row = self.row(i);
        row.to_slice()
            .expect("ndarray DataRef requires standard (row-major) layout")
    }
}

#[cfg(feature = "ndarray")]
impl super::flat::DataRef for ndarray::Array2<f32> {
    fn n(&self) -> usize {
        self.nrows()
    }
    fn d(&self) -> usize {
        self.ncols()
    }
    fn row(&self, i: usize) -> &[f32] {
        let row = self.row(i);
        row.to_slice()
            .expect("ndarray DataRef requires standard (row-major) layout")
    }
}

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

    #[test]
    fn flat_round_trip() {
        let flat = vec![1.0f32, 2.0, 3.0, 4.0];
        let vecs = flat_to_vecs(&flat, 2, 2);
        assert_eq!(vecs, vec![vec![1.0, 2.0], vec![3.0, 4.0]]);
    }

    #[test]
    #[should_panic(expected = "flat.len() must equal n * d")]
    fn flat_wrong_size() {
        flat_to_vecs(&[1.0, 2.0, 3.0], 2, 2);
    }

    #[cfg(feature = "ndarray")]
    #[test]
    fn ndarray_dataref_kmeans() {
        use ndarray::Array2;
        let arr = Array2::from_shape_vec((4, 2), vec![0.0, 0.0, 0.1, 0.1, 10.0, 10.0, 10.1, 10.1])
            .unwrap();
        let labels = crate::Kmeans::new(2)
            .with_seed(42)
            .fit_predict(&arr.view())
            .unwrap();
        assert_eq!(labels[0], labels[1]);
        assert_ne!(labels[0], labels[2]);

        // Also test owned Array2
        let labels2 = crate::Kmeans::new(2)
            .with_seed(42)
            .fit_predict(&arr)
            .unwrap();
        assert_eq!(labels2[0], labels2[1]);
        assert_ne!(labels2[0], labels2[2]);
    }
}