sproutml 0.1.1

A simple Machine Learning Library built 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
158
159
160
161
162

use serde_derive::{Serialize, Deserialize};
use rand::prelude::*;

use crate::activation::{ActivationFunction};


#[derive(Debug, Clone, Serialize, Deserialize, PartialEq)]
pub enum PaddingType {
    Same,
    Valid,
    Full
}

#[derive(Debug, Clone, Serialize, Deserialize)]
pub struct ConvParams {
    pub kernel: usize,
    pub padding_type: PaddingType,
    pub padding: usize,
    pub stride: usize,
    pub data: Vec<Vec<Vec<f64>>>, //channel > img matrix
    pub weights: Vec<Vec<Vec<f64>>>, //1 kernel > depth of kernel -> kernel weights
    pub bias: f64, //bias for each kernel not including depth
    pub outputs: Vec<Vec<Vec<f64>>>, //channel > mat
    pub inputs: Vec<Vec<Vec<f64>>>, //channel > mat
}

impl ConvParams {
    pub fn new(kernel: usize, padding_type: PaddingType, stride: usize) -> Self {
        
        let mut cp = ConvParams {
            kernel,
            padding_type,
            padding: 0,
            stride,
            data: vec![],
            weights: vec![],
            bias: 0.0,
            outputs: vec![],
            inputs: vec![],
        };
        cp
    }

    pub fn init(&mut self, activation: ActivationFunction) {
        if self.weights.len() != 0 as usize {
            return;
        }
        for i in 0..self.data.len() {
            self.weights.push(vec![vec![0.0; self.kernel]; self.kernel]);
        }

        match activation {
            ActivationFunction::Sigmoid => 
                {
                    let std_dev = (2.0 / (self.weights.len() * (self.weights[0].len() * self.weights[0].len())) as f64).sqrt();
                    let limit = (3.0 * std_dev).sqrt();
                    self.bias = 0.0;

                    for i in 0..self.weights.len() {
                        for j in 0..self.weights[i].len() {
                            for k in 0..self.weights[i][j].len() {
                                self.weights[i][j][k] = thread_rng().gen_range(-limit..limit) * std_dev;
                            }
                        }
                    }
                },
            ActivationFunction::ReLU => 
                {
                    let std_dev = (2.0 / (self.weights.len() * (self.weights[0].len() * self.weights[0].len())) as f64).sqrt();
                    let limit = (3.0 * std_dev).sqrt();
                    self.bias = 0.0;

                    for i in 0..self.weights.len() {
                        for j in 0..self.weights[i].len() {
                            for k in 0..self.weights[i][j].len() {
                                self.weights[i][j][k] = thread_rng().gen_range(-std_dev..std_dev);
                            }
                        }
                    }
                },
            ActivationFunction::TanH => 
                {
                    let std_dev = (1.0 / (self.weights.len() * (self.weights[0].len() * self.weights[0].len())) as f64).sqrt();
                    let limit = (3.0 * std_dev).sqrt();
                    self.bias = 0.0;

                    for i in 0..self.weights.len() {
                        for j in 0..self.weights[i].len() {
                            for k in 0..self.weights[i][j].len() {
                                self.weights[i][j][k] = thread_rng().gen_range(-limit..limit) * std_dev;
                            }
                        }
                    }
                },
            ActivationFunction::SoftMax => 
                {
                    let std_dev = (1.0 / (self.weights.len() * (self.weights[0].len() * self.weights[0].len())) as f64).sqrt();
                    let limit = (3.0 * std_dev).sqrt();
                    self.bias = 0.0;

                    for i in 0..self.weights.len() {
                        for j in 0..self.weights[i].len() {
                            for k in 0..self.weights[i][j].len() {
                                self.weights[i][j][k] = thread_rng().gen_range(-limit..limit) * std_dev;
                            }
                        }
                    }
                },
        }
    }

    pub fn add_padding(&mut self, activation: ActivationFunction) {
        if self.padding_type == PaddingType::Valid {
            self.data = self.inputs.clone();
            self.init(activation);
            return;
        }
        let mut padded_image = vec![];
        let mut padding = (self.kernel - 1) / 2;

        if self.padding_type == PaddingType::Full {
            padding = self.kernel - 1;
        }

        self.padding = padding;
        let height = self.inputs[0].len();
        let width = self.inputs[0][0].len();
        let padded_height = height + 2 * self.padding;
        let padded_width = width + 2 * self.padding;
        
        for i in 0..self.inputs.len() {
            padded_image.push(vec![vec![0.0; padded_width]; padded_height]);
        }

        for i in 0..padded_image.len() {
            for j in 0..height {
                for k in 0..width {
                    padded_image[i][j + self.padding][k + self.padding] = self.inputs[i][j][k];
                }
            }
        }

        self.data = padded_image;
        self.init(activation);
    }

    pub fn get_output_dims(&self) -> [usize; 2] {
        let height = self.data[0].len();
        let width = self.data[0][0].len();

        let out_width = (width - self.kernel) / self.stride + 1;
        let out_height = (height - self.kernel) / self.stride + 1;

        [out_width, out_height]
    }
    
    pub fn print_kernels(&self) {
        println!("--------------------------\nKernel Dimensions: {} x {}", self.kernel, self.kernel);
        println!("Weights: \n{:#?}", self.weights);
        println!("Biases: \n{:#?}\n-------------------------------", self.bias);
    }
}