automl 0.3.0

Automated machine learning for classification and regression
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

//! Feature engineering helpers.
//!
//! This module provides pure functions to expand feature sets, keeping
//! transformations separate from model logic.

use itertools::Itertools;
use smartcore::linalg::basic::arrays::{Array, Array2};
use smartcore::linalg::basic::matrix::DenseMatrix;
use smartcore::numbers::floatnum::FloatNumber;
use smartcore::numbers::realnum::RealNumber;
use std::error::Error;
use std::fmt::{self, Display, Formatter};

use crate::utils::math::elementwise_multiply;

/// Errors that can occur during feature construction.
#[derive(Clone, Copy, Debug, PartialEq, Eq)]
pub enum FeatureError {
    /// Failed to build an intermediate matrix.
    MatrixCreationFailed,
    /// The input matrix has zero height or width.
    InvalidInputDimensions,
}

impl Display for FeatureError {
    fn fmt(&self, f: &mut Formatter<'_>) -> fmt::Result {
        match self {
            Self::MatrixCreationFailed => write!(f, "cannot create matrix"),
            Self::InvalidInputDimensions => write!(f, "input matrix has invalid dimensions"),
        }
    }
}

impl Error for FeatureError {}

/// Generate pairwise interaction features by multiplying each combination of
/// columns.
///
/// # Arguments
///
/// * `x` - Input feature matrix.
///
/// # Examples
///
/// ```
/// use automl::utils::interaction_features;
/// use smartcore::linalg::basic::arrays::Array;
/// use smartcore::linalg::basic::matrix::DenseMatrix;
/// use automl::utils::FeatureError;
/// # fn main() -> Result<(), FeatureError> {
/// let x = DenseMatrix::from_2d_vec(&vec![vec![1.0, 2.0], vec![3.0, 4.0]]).unwrap();
/// let augmented = interaction_features(x)?;
/// assert_eq!(augmented.shape(), (2, 3));
/// # Ok(())
/// # }
/// ```
///
/// # Errors
///
/// Returns an error if the input matrix is empty or a new feature column
/// cannot be created.
pub fn interaction_features<INPUT, InputArray>(
    mut x: InputArray,
) -> Result<InputArray, FeatureError>
where
    INPUT: RealNumber + FloatNumber,
    InputArray: Clone + Array<INPUT, (usize, usize)> + Array2<INPUT>,
{
    let (height, width) = x.shape();
    if height == 0 || width == 0 {
        return Err(FeatureError::InvalidInputDimensions);
    }
    for column_1 in 0..width {
        for column_2 in (column_1 + 1)..width {
            let col1: Vec<INPUT> = (0..height)
                .map(|idx| *x.get_col(column_1).get(idx))
                .collect();
            let col2: Vec<INPUT> = (0..height)
                .map(|idx| *x.get_col(column_2).get(idx))
                .collect();
            let feature = elementwise_multiply(&col1, &col2);
            let new_column = DenseMatrix::from_2d_vec(&vec![feature; 1])
                .map_err(|_| FeatureError::MatrixCreationFailed)?
                .transpose();
            x = x.h_stack(&new_column);
        }
    }
    Ok(x)
}

/// Generate polynomial features up to a given order.
///
/// # Arguments
///
/// * `x` - Input feature matrix.
/// * `order` - Highest polynomial order to generate.
///
/// # Examples
///
/// ```
/// use automl::utils::polynomial_features;
/// use smartcore::linalg::basic::arrays::Array;
/// use smartcore::linalg::basic::matrix::DenseMatrix;
/// use automl::utils::FeatureError;
/// # fn main() -> Result<(), FeatureError> {
/// let x = DenseMatrix::from_2d_vec(&vec![vec![1.0, 2.0], vec![3.0, 4.0]]).unwrap();
/// let augmented = polynomial_features(x, 2)?;
/// assert_eq!(augmented.shape(), (2, 5));
/// # Ok(())
/// # }
/// ```
///
/// # Errors
///
/// Returns an error if the input matrix is empty or a new feature column
/// cannot be created.
pub fn polynomial_features<INPUT, InputArray>(
    mut x: InputArray,
    order: usize,
) -> Result<InputArray, FeatureError>
where
    INPUT: RealNumber + FloatNumber,
    InputArray: Clone + Array<INPUT, (usize, usize)> + Array2<INPUT>,
{
    let (height, width) = x.shape();
    if height == 0 || width == 0 {
        return Err(FeatureError::InvalidInputDimensions);
    }
    for n in 2..=order {
        let combinations = (0..width).combinations_with_replacement(n);
        for combo in combinations {
            let mut feature: Vec<INPUT> = vec![INPUT::one(); height];
            for column in combo {
                let col: Vec<INPUT> = (0..height).map(|idx| *x.get_col(column).get(idx)).collect();
                feature = elementwise_multiply(&col, &feature);
            }
            let new_column = DenseMatrix::from_2d_vec(&vec![feature; 1])
                .map_err(|_| FeatureError::MatrixCreationFailed)?
                .transpose();
            x = x.h_stack(&new_column);
        }
    }
    Ok(x)
}