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

//! Functions to compute the k-nearest neighbours.
extern crate num;

use self::num::traits::Float;
use matrix::*;
use vectors::group;


/*
use distance::Euclid;

pub struct Knn<T> {
    distance: Fn(&[T], &[T]) -> T
}

impl Knn<f64> {

    pub fn new() -> Knn<f64> {
        Knn {
            distance: |x, y| Euclid::compute(x, y).unwrap(),
        }
    }
}
*/

/// Search the k nearest neighbours for the given example.
pub fn scan<D, T: Float>(m: &Matrix<T>, example: &[T], k: usize, df: D) -> Option<Vec<usize>>
    where D : Fn(&[T], &[T]) -> T {

    if example.len() != m.cols() {
        return None;
    }

    let mut near: Vec<(usize, T)> = Vec::with_capacity(k);

    for (idx, row) in m.row_iter().enumerate() {
        let d = df(row, example);

        // search the first neighbour for which the distance is larger
        // than the distance to the current example and insert it at that
        // position
        let p = near.iter().position(|&(_, val)| val > d);
        match p {
            Some(pos) => {
                near.insert(pos, (idx, d));
                if near.len() > k {
                    near.pop();
                }
            }
            _ => {
                if idx < k {
                    near.push((idx, d))
                }
            }
        }
    }

    Some(near.iter().map(|&(idx, _)| idx.clone()).collect())
}

pub fn classify<T, L, D>(m: &Matrix<T>, labels: &Vec<L>, example: &[T], k: usize, df: D) -> L
    where T: Float, L: Clone + Ord, D: Fn(&[T], &[T]) -> T {

    let idx = scan(&m, example, k, df).unwrap();

    let mut targets: Vec<L> = idx.iter().map(|pos| labels.get(*pos).unwrap()).cloned().collect();
    targets.sort_by(|a, b| a.cmp(&b));
    let mut r = group(&targets);
    r.sort_by(|a, b| a.1.cmp(&b.1));

    r.last().unwrap().0.clone()
}


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

    #[test]
    fn test_knn_classify() {

        let m = mat![
            1.0, 2.0;
            1.1, 2.1;
            2.0, 3.0;
            0.9, 1.9;
            2.1, 2.9
        ];

        let labels = vec![1, 2, 2, 1, 2];
        let target = classify(&m, &labels, &[1.3, 2.0], 3, |x, y| Euclid::compute(x, y).unwrap());
        assert_eq!(target, 1);
    }

    #[test]
    fn test_scan() {

        let mut m = mat![
            1.0, 2.0;
            2.0, 2.0;
            3.0, 3.0
        ];

        let a = scan(&m, &[1.0, 1.0, 2.0], 1, |x, y| Euclid::compute(x, y).unwrap());
        assert!(a.is_none());

        let mut label = scan(&m, &[1.0, 1.0], 1, |x, y| Euclid::compute(x, y).unwrap()).unwrap();
        assert_eq!(label, vec![0]);

        label = scan(&m, &[1.0, 2.0], 1, |x, y| Euclid::compute(x, y).unwrap()).unwrap();
        assert_eq!(label, vec![0]);

        label = scan(&m, &[2.0, 2.2], 1, |x, y| Euclid::compute(x, y).unwrap()).unwrap();
        assert_eq!(label, vec![1]);

        label = scan(&m, &[5.0, 6.0], 1, |x, y| Euclid::compute(x, y).unwrap()).unwrap();
        assert_eq!(label, vec![2]);

        // ---------------

        m = mat![
            1.0, 2.0;
            1.3, 1.8;
            1.2, 2.1;
            2.0, 2.0
        ];

        label = scan(&m, &[1.1, 2.0], 1, |x, y| Euclid::compute(x, y).unwrap()).unwrap();
        assert_eq!(label, vec![0]);
        label = scan(&m, &[1.1, 2.0], 2, |x, y| Euclid::compute(x, y).unwrap()).unwrap();
        assert_eq!(label, vec![0, 2]);
        label = scan(&m, &[1.1, 2.0], 3, |x, y| Euclid::compute(x, y).unwrap()).unwrap();
        assert_eq!(label, vec![0, 2, 1]);
        label = scan(&m, &[1.1, 2.0], 4, |x, y| Euclid::compute(x, y).unwrap()).unwrap();
        assert_eq!(label, vec![0, 2, 1, 3]);
    }
}