datarust 0.7.0

Scikit-learn-style preprocessing and classical ML in 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

use datarust::ensemble::{RandomForestClassifier, RandomForestRegressor};
use datarust::traits::{PredictProba, Predictor};
use datarust::{Matrix, MaxFeatures, ParamValue, Params};

fn fixture() -> Matrix {
    Matrix::new(vec![
        vec![0.0, 0.0, 1.0],
        vec![0.0, 1.0, 0.0],
        vec![1.0, 0.0, 0.0],
        vec![1.0, 1.0, 1.0],
        vec![2.0, 0.0, 1.0],
        vec![2.0, 1.0, 0.0],
        vec![3.0, 0.0, 0.0],
        vec![3.0, 1.0, 1.0],
    ])
    .unwrap()
}

#[test]
fn classifier_matches_sklearn_reference_predictions_probabilities_and_importances() {
    let x = fixture();
    let y = vec![0.0, 1.0, 1.0, 0.0, 0.0, 1.0, 1.0, 0.0];
    let mut model = RandomForestClassifier::new()
        .with_n_estimators(5)
        .with_max_depth(Some(3))
        .with_max_features(MaxFeatures::Count(2))
        .with_random_state(13);
    model.fit(&x, &y).unwrap();

    assert_eq!(model.predict(&x).unwrap(), y);
    assert_eq!(
        model.predict_proba(&x).unwrap().as_slice(),
        &[0.8, 0.2, 0.2, 0.8, 0.0, 1.0, 1.0, 0.0, 1.0, 0.0, 0.2, 0.8, 0.4, 0.6, 1.0, 0.0,]
    );
    let expected_importances: [f64; 3] =
        [0.13142857142857145, 0.04571428571428573, 0.8228571428571428];
    for (actual, expected) in model
        .feature_importances()
        .unwrap()
        .iter()
        .zip(expected_importances)
    {
        assert_eq!(actual.to_bits(), expected.to_bits());
    }
}

#[test]
fn regressor_matches_sklearn_reference_predictions_importances_and_bootstrap_indices() {
    let x = fixture();
    let y = vec![0.5, 2.0, 1.0, 2.5, 1.5, 3.0, 2.0, 3.5];
    let mut model = RandomForestRegressor::new()
        .with_n_estimators(5)
        .with_max_depth(Some(3))
        .with_max_features(MaxFeatures::Count(2))
        .with_random_state(13);
    model.fit(&x, &y).unwrap();

    let expected_predictions: [f64; 8] = [0.9, 2.2, 1.0, 2.4, 1.4, 2.8, 1.5, 3.5];
    for (actual, expected) in model.predict(&x).unwrap().iter().zip(expected_predictions) {
        assert_eq!(actual.to_bits(), expected.to_bits());
    }
    let expected_importances: [f64; 3] =
        [0.24508660036871355, 0.6780638747828597, 0.07684952484842672];
    for (actual, expected) in model
        .feature_importances()
        .unwrap()
        .iter()
        .zip(expected_importances)
    {
        assert_eq!(actual.to_bits(), expected.to_bits());
    }
    assert_eq!(
        model.estimators_samples(),
        vec![
            vec![7, 4, 4, 3, 1, 2, 7, 6],
            vec![3, 6, 3, 2, 1, 7, 5, 1],
            vec![3, 5, 7, 2, 2, 4, 5, 0],
            vec![0, 1, 7, 5, 3, 0, 4, 2],
            vec![1, 7, 2, 4, 5, 4, 5, 1],
        ]
    );
}

#[test]
fn bootstrap_false_uses_each_training_row_once_and_params_round_trip() {
    let x = fixture();
    let y = vec![0.5, 2.0, 1.0, 2.5, 1.5, 3.0, 2.0, 3.5];
    let mut model = RandomForestRegressor::new()
        .with_n_estimators(3)
        .with_bootstrap(false)
        .with_max_features(MaxFeatures::Count(2));
    model.fit(&x, &y).unwrap();
    assert!(model
        .estimators_samples()
        .iter()
        .all(|samples| samples == &(0..x.nrows()).collect::<Vec<_>>()));
    for params in model.get_params() {
        model.set_params(params.0, params.1).unwrap();
    }
    // Params encode max_features as 0=all, 1=sqrt, 2=log2, k+2=Count(k);
    // Int(4) therefore round-trips Count(2) set above.
    assert!(model
        .get_params()
        .contains(&("max_features", ParamValue::Int(4))));
}

#[test]
fn classifier_accepts_one_class_and_returns_one_probability_column() {
    let x = fixture();
    let y = vec![7.0; 8];
    let mut model = RandomForestClassifier::new()
        .with_n_estimators(3)
        .with_random_state(5);
    model.fit(&x, &y).unwrap();
    assert_eq!(model.classes(), Some(&[7.0][..]));
    let proba = model.predict_proba(&x).unwrap();
    assert_eq!(proba.ncols(), 1);
    assert!(proba.as_slice().iter().all(|&value| value == 1.0));
    assert_eq!(model.predict(&x).unwrap(), y);
}

#[test]
fn classifier_is_deterministic_and_preserves_external_class_labels() {
    let x = fixture();
    let y = vec![10.0, 20.0, 20.0, 10.0, 10.0, 20.0, 20.0, 10.0];
    let build = || {
        let mut model = RandomForestClassifier::new()
            .with_n_estimators(7)
            .with_max_features(MaxFeatures::Count(1))
            .with_random_state(99);
        model.fit(&x, &y).unwrap();
        model
    };
    let a = build();
    let b = build();
    assert_eq!(a.classes(), Some(&[10.0, 20.0][..]));
    assert_eq!(a.predict(&x).unwrap(), b.predict(&x).unwrap());
    assert_eq!(a.predict_proba(&x).unwrap(), b.predict_proba(&x).unwrap());
}

#[test]
fn classifier_rejects_regression_criterion_at_fit() {
    use datarust::tree::Criterion;
    use datarust::DatarustError;

    // Regression: the forest used to fit "successfully" and only fail on
    // the first predict. `fit` must reject the regression-only criterion
    // before any tree is built.
    let x = fixture();
    let y = vec![0.0, 1.0, 1.0, 0.0, 0.0, 1.0, 1.0, 0.0];
    let mut model = RandomForestClassifier::new()
        .with_n_estimators(3)
        .with_criterion(Criterion::SquaredError);
    let err = model.fit(&x, &y).unwrap_err();
    assert!(matches!(err, DatarustError::InvalidInput(_)), "got {err:?}");
    assert_eq!(model.n_estimators(), 0);
}