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

use crate::prelude::*;

/// Normalizes `t` to have mean `0.0` and stddev `1.0` along the `I` dimension of `T`. `epsilon` is passed to [std_axis()].
/// Computes `(t - t.mean(I)) / t.std(I, epsilon)`.
///
/// **Related functions:** [mean_axis()], [std_axis()], [var_axis()]
///
/// # Examples
/// ```rust
/// # use dfdx::prelude::*;
/// let a = Tensor1D::new([-2.0, -1.0, 0.0, 5.0, 2.0]);
/// let r = a.normalize_axis::<-1>(1e-5);
/// assert!(r.clone().mean_axis::<-1>().data().abs() < 1e-6);
/// assert!((r.clone().std_axis::<-1>(0.0).data() - 1.0).abs() < 1e-6);
/// ```
pub fn normalize_axis<T, const I: isize>(t: T, epsilon: T::Dtype) -> T
where
    T: Reduce1<I>,
    T::Array: HasAxis<I>,
{
    let (t, tape) = t.split_tape();
    let (std, tape) = std_axis::<T, I>(t.duplicate().put_tape(tape), epsilon)
        .broadcast1()
        .split_tape();
    let (mean, tape) = mean_axis::<T, I>(t.duplicate().put_tape(tape))
        .broadcast1()
        .split_tape();
    let centered = sub(t.put_tape(tape), &mean);
    div(centered, &std)
}

macro_rules! tensor_impl {
    ($typename:ident, [$($Vs:tt),*]) => {
impl<$(const $Vs: usize, )* H: Tape> $typename<$($Vs, )* H>
{
    /// Calls [normalize_axis()] on `self`.
    pub fn normalize_axis<const I: isize>(self, epsilon: f32) -> Self
    where
        Self: Reduce1<I>,
        <Self as HasArrayType>::Array: HasAxis<I>,
    {
        normalize_axis::<Self, I>(self, epsilon)
    }
}
    };
}

tensor_impl!(Tensor0D, []);
tensor_impl!(Tensor1D, [M]);
tensor_impl!(Tensor2D, [M, N]);
tensor_impl!(Tensor3D, [M, N, O]);
tensor_impl!(Tensor4D, [M, N, O, P]);

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

    #[test]
    fn test_0d_normalize_axis_last() {
        let a = Tensor0D::new(10.0);
        let r = a.trace().normalize_axis::<-1>(1e-5);
        assert_eq!(r.data(), &0.0);
        let gradients = r.backward();
        assert_eq!(gradients.ref_gradient(&a), &0.0);
    }

    #[test]
    fn test_1d_normalize_axis_last() {
        let a = Tensor1D::new([-2.0, 0.0, 5.0]);
        let r = a.trace().normalize_axis::<-1>(1e-5);
        assert_eq!(r.data(), &[-1.0190487, -0.3396829, 1.3587316]);
        // NOTE: .exp() so we can make sure normalize is using result grad properly
        let gradients = r.exp().mean().backward();
        assert_eq!(
            gradients.ref_gradient(&a),
            &[0.033410847, -0.04677555, 0.013364702]
        );
    }

    #[test]
    fn test_2d_normalize_axis_last() {
        let a: Tensor2D<2, 3> = Tensor2D::new([[-2.0, 0.0, 5.0], [1.0, 2.0, 3.0]]);
        let r = a.trace().normalize_axis::<-1>(1e-5);
        assert_eq!(
            r.data(),
            &[
                [-1.0190487, -0.3396829, 1.3587316],
                [-1.2247356, 0.0, 1.2247356]
            ]
        );
        let gradients = r.exp().mean().backward();
        assert_eq!(
            gradients.ref_gradient(&a),
            &[
                [0.016705424, -0.023387775, 0.006682351],
                [0.05773133, -0.11547226, 0.057740927]
            ]
        );
    }

    #[test]
    fn test_2d_normalize_axis_first() {
        let a: Tensor2D<3, 2> = Tensor2D::new([[-2.0, 0.0], [1.0, 2.0], [4.0, 5.0]]);
        let r = a.trace().normalize_axis::<0>(1e-5);
        assert_eq!(
            r.data(),
            &[
                [-1.2247438, -1.1355485],
                [0.0, -0.16222118],
                [1.2247438, 1.2977698],
            ]
        );
        let gradients = r.exp().mean().backward();
        assert_eq!(
            gradients.ref_gradient(&a),
            &[
                [0.019245632, 0.025835907],
                [-0.038491584, -0.043060362],
                [0.019245982, 0.017224466],
            ]
        );
    }

    #[test]
    fn test_3d_normalize_axis_last() {
        let a: Tensor3D<4, 2, 3> = Tensor3D::ones();
        let r = a.trace().normalize_axis::<-1>(1e-5);
        assert_eq!(r.data(), &[[[0.0; 3]; 2]; 4]);
        let gradients = r.exp().mean().backward();
        assert_eq!(gradients.ref_gradient(&a), &[[[0.0; 3]; 2]; 4]);
    }
}