esp-idf-unda 0.1.2

Embedded neural network forward feeding for precompiled models created by the unda crate
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

use std::f32::consts::E;

use super::{matrix::Matrix, input::Input};

#[derive(Clone)]
pub struct Activation<'a>{
    pub function: &'a dyn Fn(f32) -> f32,
    pub derivative: &'a dyn Fn(f32) -> f32
}

#[derive(Clone, Copy)]
pub enum Activations{
    SIGMOID,
    TANH,
    RELU,
    LEAKYRELU,
    SOFTMAX,
    ELU(f32)
}
impl Activations{
    fn get_function(&self) -> Option<Activation>{
        return match self{
            Activations::SIGMOID => Some(SIGMOID),
            Activations::TANH => Some(TANH),
            Activations::RELU => Some(RELU),
            Activations::LEAKYRELU => Some(LEAKY_RELU),
            _ => None
        };
    }
    pub fn apply_fn(&self, mut data: Matrix) -> Matrix {
        match self{
            Activations::SIGMOID | Activations::TANH | Activations::RELU | Activations::LEAKYRELU => {
                return data.map(self.get_function().unwrap().function);
            },
            Activations::SOFTMAX => { 
                let exp_logits: Vec<f32> = data.to_param()
                    .iter()
                    .map(|&x| x.exp()).collect();
                let sum_exp: f32 = exp_logits.iter().sum();
                return Matrix::from_sized(exp_logits.iter().map(|x| x / sum_exp).collect::<Vec<f32>>(), data.rows, data.columns)
            },
            Activations::ELU(alpha) => {
                let data_elu = data.to_param()
                    .iter()
                    .map(|&x| elu(*alpha, x)).collect();
                return Matrix::from_sized(data_elu, data.rows, data.columns);
            }

        };
    }
    pub fn apply_derivative(&self, mut data: Matrix) -> Matrix {
        match self{
            Activations::SIGMOID | Activations::TANH | Activations::RELU | Activations::LEAKYRELU => {
                return data.map(self.get_function().unwrap().derivative);
            },
            Activations::SOFTMAX => { 
                let softmax_output = data.to_param()
                    .iter()
                    .zip(data.to_param().iter().map(|&x| 1.0 - x))
                    .map(|(s,ds)| s * ds)
                    .collect();
                return Matrix::from_sized(softmax_output, data.rows, data.columns);
            },
            Activations::ELU(alpha) => {
                let data_elu = data.to_param()
                    .iter()
                    .map(|&x| d_elu(*alpha, x)).collect();
                return Matrix::from_sized(data_elu, data.rows, data.columns);
            }
        };
    }
}


const SIGMOID: Activation = Activation {
    function: &|x| {
        let res = 1.0 / (1.0 + E.powf(-x));
        return res;
    },
    derivative: &|x| x * (1.0 - x)
};

const TANH: Activation = Activation {
    function: &|x| {
        let res = f32::tanh(x);
        return res;
    },
    derivative: &|x| 1.0 - f32::tanh(x).powf(2.0)
};

const RELU: Activation = Activation {
    function: &|x| {
        let res = x.max(0.0);
        return res;
    },
    derivative: &|x| {
        if x.max(0.0) == x {
            return 1.0;
        }
        return 0.0;
    }
};

const LEAKY_RELU: Activation = Activation{
    function: &|x| {
        if x > 0.0{
            return x;
        }
        return 0.001 * x;
    },
    derivative: &|x| {
        if x.max(0.0) == x{
            return 1.0
        }
        return 0.001;
    }
};

fn d_elu(alpha: f32, x: f32) -> f32 {
     if x > 0.0 {
        return 1.0;
    }
    return alpha * E.powf(x);
   
}

fn elu(alpha: f32, x: f32) -> f32 {
    if x.max(0.0) == x{
        return x;
    }
    return alpha * (E.powf(x) - 1.0);
}