unda 0.2.2

General purpose machine learning crate for neural network development and analysis
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

use super::{matrix::Matrix, matrix3d::Matrix3D};
///Dynamic trait for input objects as they go in and travel through a neural network
///As long as your data type implements Input, it can be used through convolutional, dense,
///and any other layer type without the need for flattening layers or shape specifiers.
///
///Dense layers always use to_param for single input and to_param_2d to generate matrices through
///the network
///Convolutional layers always use to_param_2d for single inputs and to_param_3d for through-layer
///traveling
pub trait Input: Send + Sync{
    ///Flattens input data into a one dimensional context
    ///Most commonly used by dense layers and output layers
    fn to_param(&self) -> Vec<f32>;
    ///Flattens(or extends) data into a two dimensional context
    ///Flattens if data is a higher order, extends if its a lower order
    fn to_param_2d(&self) -> Vec<Vec<f32>>;
    ///Flattens(or extends) data into a two dimensional context
    ///Flattens if data is a higher order, extends if its a lower order
    fn to_param_3d(&self) -> Vec<Vec<Vec<f32>>>;
    ///Returns the underlying shape the data has when it is not being morphed or shaped by the
    ///param methods
    fn shape(&self) -> (usize, usize, usize);
    ///Wrapper method for boxing an input up
    fn to_box(&self) -> Box<dyn Input>;
}

///Input implementation for 1 dimensional floating point vectors
impl Input for Vec<f32>{
    fn to_param(&self) -> Vec<f32>{
        self.clone()
    }
    fn to_param_2d(&self) -> Vec<Vec<f32>> {
        vec![self.clone()]
    }
    fn to_param_3d(&self) -> Vec<Vec<Vec<f32>>> {
        vec![vec![self.clone()]]
    }
    fn shape(&self) -> (usize, usize, usize) {
        (self.len(), 1, 1)
    }
    fn to_box(&self) -> Box<dyn Input> {
        Box::new(self.to_param())
    }
}
///Input implementation for 2 dimensional floating point vectors(matrix data)
impl Input for Vec<Vec<f32>> {
    fn to_param(&self) -> Vec<f32> {
        self.clone().into_iter().flatten().collect::<Vec<f32>>()
    }
    fn to_param_2d(&self) -> Vec<Vec<f32>> {
        self.clone()
    }
    fn to_param_3d(&self) -> Vec<Vec<Vec<f32>>> {
        vec![self.clone()]
    }
    fn shape(&self) -> (usize, usize, usize) {
        (self.len(), self[0].len(), 1)
    }
    fn to_box(&self) -> Box<dyn Input> {
        Box::new(self.to_param_2d())
    }
}

///Input implementation for 3 dimensional floating point vectors(3d matrix data or can be viewed as
///convolutional outputs)
impl Input for Vec<Vec<Vec<f32>>> {
    fn to_param(&self) -> Vec<f32> {
        self.clone().into_iter().flatten().flatten().collect::<Vec<f32>>()
    }
    fn to_param_2d(&self) -> Vec<Vec<f32>> {
        self.clone().into_iter().flatten().collect::<Vec<Vec<f32>>>()
    }
    fn to_param_3d(&self) -> Vec<Vec<Vec<f32>>> {
        self.clone()
    }
    fn shape(&self) -> (usize, usize, usize) {
        (self[0][0].len(), self[0].len(), self.len())
    }
    fn to_box(&self) -> Box<dyn Input> {
        Box::new(self.to_param_3d())
    }
}
impl Input for Matrix {
    fn to_param(&self) -> Vec<f32> {
        self.data.clone().into_iter().flatten().collect::<Vec<f32>>()
    }
    fn to_param_2d(&self) -> Vec<Vec<f32>> {
        self.data.clone()
    }
    fn to_param_3d(&self) -> Vec<Vec<Vec<f32>>> {
        vec![self.data.clone()]
    }
    fn shape(&self) -> (usize, usize, usize) {
        (self.rows, self.columns, 1)
    }
    fn to_box(&self) -> Box<dyn Input> {
        Box::new(self.to_param_2d())
    }
}

impl Input for Matrix3D {
    fn to_param(&self) -> Vec<f32> {
        self.data.clone().into_iter().flatten().flatten().collect::<Vec<f32>>()
    }
    fn to_param_2d(&self) -> Vec<Vec<f32>> {
        self.data.clone().into_iter().flatten().collect::<Vec<Vec<f32>>>()
    }
    fn to_param_3d(&self) -> Vec<Vec<Vec<f32>>> {
        self.data.clone()
    }
    fn shape(&self) -> (usize, usize, usize) {
        (self.rows, self.columns, self.layers)
    }
    fn to_box(&self) -> Box<dyn Input> {
        Box::new(self.to_param_3d())
    }
}

impl Into<Box<dyn Input>> for Vec<f32> {
    fn into(self) -> Box<dyn Input> {
        Box::new(self)
    }
}