linfa 0.8.1

A Machine Learning framework for 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

use crate::Float;
use ndarray::{ArrayBase, Data, Dimension, OwnedRepr, ViewRepr};

/// Add the Lapack bound to the floating point of a dataset
///
/// This helper trait is introduced to avoid leaking `Lapack + Scalar` bounds to the outside which
/// causes ambiguities when calling functions like `abs` for `num_traits::Float` and
/// `Cauchy::Scalar`. We are only using real values here, but the LAPACK routines
/// require that `Cauchy::Scalar` is implemented.
pub trait WithLapack<D: Data + WithLapackData, I: Dimension> {
    fn with_lapack(self) -> ArrayBase<D::D, I>;
}

impl<F: Float, D, I> WithLapack<D, I> for ArrayBase<D, I>
where
    D: Data<Elem = F> + WithLapackData,
    I: Dimension,
{
    fn with_lapack(self) -> ArrayBase<D::D, I> {
        D::with_lapack(self)
    }
}

/// Remove the Lapack bound to the floating point of a dataset
///
/// This helper trait is introduced to avoid leaking `Lapack + Scalar` bounds to the outside which
/// causes ambiguities when calling functions like `abs` for `num_traits::Float` and
/// `Cauchy::Scalar`. We are only using real values here, but the LAPACK routines
/// require that `Cauchy::Scalar` is implemented.
pub trait WithoutLapack<F: Float, D: Data + WithoutLapackData<F>, I: Dimension> {
    fn without_lapack(self) -> ArrayBase<D::D, I>;
}

impl<F: Float, D, I> WithoutLapack<F, D, I> for ArrayBase<D, I>
where
    D: Data<Elem = F::Lapack> + WithoutLapackData<F>,
    I: Dimension,
{
    fn without_lapack(self) -> ArrayBase<D::D, I> {
        D::without_lapack(self)
    }
}

unsafe fn transmute<A, B>(a: A) -> B {
    let b = std::ptr::read(&a as *const A as *const B);
    std::mem::forget(a);

    b
}

pub trait WithLapackData
where
    Self: Data,
{
    type D: Data;

    /// Add trait bound `Lapack` and `Scalar` to NdArray's floating point
    ///
    /// This is safe, because only implemented for D == Self
    fn with_lapack<I>(x: ArrayBase<Self, I>) -> ArrayBase<Self::D, I>
    where
        I: Dimension,
    {
        unsafe { transmute(x) }
    }
}

impl<F: Float> WithLapackData for OwnedRepr<F> {
    type D = OwnedRepr<F::Lapack>;
}

impl<'a, F: Float> WithLapackData for ViewRepr<&'a F> {
    type D = ViewRepr<&'a F::Lapack>;
}

impl<'a, F: Float> WithLapackData for ViewRepr<&'a mut F> {
    type D = ViewRepr<&'a mut F::Lapack>;
}

pub trait WithoutLapackData<F: Float>
where
    Self: Data,
{
    type D: Data<Elem = F>;

    /// Add trait bound `Lapack` and `Scalar` to NdArray's floating point
    ///
    /// This is safe, because only implemented for D == Self
    fn without_lapack<I>(x: ArrayBase<Self, I>) -> ArrayBase<Self::D, I>
    where
        I: Dimension,
    {
        unsafe { transmute(x) }
    }
}

impl<F: Float> WithoutLapackData<F> for OwnedRepr<F::Lapack> {
    type D = OwnedRepr<F>;
}

impl<'a, F: Float> WithoutLapackData<F> for ViewRepr<&'a F::Lapack> {
    type D = ViewRepr<&'a F>;
}

impl<'a, F: Float> WithoutLapackData<F> for ViewRepr<&'a mut F::Lapack> {
    type D = ViewRepr<&'a mut F>;
}

#[cfg(test)]
mod tests {
    use ndarray::Array2;
    #[cfg(feature = "ndarray-linalg")]
    use ndarray_linalg::eig::*;

    use super::{WithLapack, WithoutLapack};

    #[test]
    fn memory_check() {
        let a: Array2<f32> = Array2::zeros((20, 20));
        let a: Array2<f32> = a.with_lapack();

        assert_eq!(a.shape(), &[20, 20]);

        let b: Array2<f32> = a.clone().without_lapack();

        assert_eq!(b, a);
    }

    #[cfg(feature = "ndarray-linalg")]
    #[test]
    fn lapack_exists() {
        let a: Array2<f32> = Array2::zeros((4, 4));
        let a: Array2<f32> = a.with_lapack();

        let (_a, _b) = a.eig().unwrap();
    }
}