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
|
/*
Copyright (c) 2016, Taiga Nomi
All rights reserved.
Redistribution and use in source and binary forms, with or without
modification, are permitted provided that the following conditions are met:
* Redistributions of source code must retain the above copyright
notice, this list of conditions and the following disclaimer.
* Redistributions in binary form must reproduce the above copyright
notice, this list of conditions and the following disclaimer in the
documentation and/or other materials provided with the distribution.
* Neither the name of the <organization> nor the
names of its contributors may be used to endorse or promote products
derived from this software without specific prior written permission.
THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY
EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE FOR ANY
DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
(INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND
ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
(INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*/
#pragma once
#include "tiny_dnn/util/util.h"
#include "tiny_dnn/layers/layer.h"
#include <cmath>
namespace tiny_dnn {
/**
* element-wise pow: ```y = scale*x^factor```
**/
class power_layer : public layer {
public:
typedef layer Base;
/**
* @param in_shape [in] shape of input tensor
* @param factor [in] floating-point number that specifies a power
* @param scale [in] scale factor for additional multiply
*/
power_layer(const shape3d& in_shape, float_t factor, float_t scale=1.0f)
: layer({ vector_type::data }, { vector_type::data }),
in_shape_(in_shape), factor_(factor), scale_(scale) {
}
/**
* @param prev_layer [in] previous layer to be connected
* @param factor [in] floating-point number that specifies a power
* @param scale [in] scale factor for additional multiply
*/
power_layer(const layer& prev_layer, float_t factor, float_t scale=1.0f)
: layer({ vector_type::data }, { vector_type::data }),
in_shape_(prev_layer.out_shape()[0]), factor_(factor), scale_(scale) {
}
std::string layer_type() const override {
return "power";
}
std::vector<shape3d> in_shape() const override {
return {in_shape_};
}
std::vector<shape3d> out_shape() const override {
return {in_shape_};
}
void forward_propagation(const std::vector<tensor_t*>& in_data,
std::vector<tensor_t*>& out_data) override {
const tensor_t& x = *in_data[0];
tensor_t& y = *out_data[0];
for (serial_size_t i = 0; i < x.size(); i++) {
std::transform(x[i].begin(), x[i].end(), y[i].begin(), [=](float_t x) {
return scale_*std::pow(x, factor_);
});
}
}
void back_propagation(const std::vector<tensor_t*>& in_data,
const std::vector<tensor_t*>& out_data,
std::vector<tensor_t*>& out_grad,
std::vector<tensor_t*>& in_grad) override {
tensor_t& dx = *in_grad[0];
const tensor_t& dy = *out_grad[0];
const tensor_t& x = *in_data[0];
const tensor_t& y = *out_data[0];
for (serial_size_t i = 0; i < x.size(); i++) {
for (serial_size_t j = 0; j < x[i].size(); j++) {
// f(x) = (scale*x)^factor
// ->
// dx = dy * df(x)
// = dy * scale * factor * (scale * x)^(factor - 1)
// = dy * scale * factor * (scale * x)^factor * (scale * x)^(-1)
// = dy * factor * y / x
if (std::abs(x[i][j]) > 1e-10) {
dx[i][j] = dy[i][j] * factor_ * y[i][j] / x[i][j];
}
else {
dx[i][j] = dy[i][j] * scale_ * factor_ * std::pow(x[i][j], factor_ - 1.0f);
}
}
}
}
template <class Archive>
static void load_and_construct(Archive & ar, cereal::construct<power_layer> & construct) {
shape3d in_shape;
float_t factor;
float_t scale(1.0f);
ar(cereal::make_nvp("in_size", in_shape), cereal::make_nvp("factor", factor), cereal::make_nvp("scale", scale));
construct(in_shape, factor, scale);
}
template <class Archive>
void serialize(Archive & ar) {
layer::serialize_prolog(ar);
ar(cereal::make_nvp("in_size", in_shape_), cereal::make_nvp("factor", factor_), cereal::make_nvp("scale", scale_));
}
float_t factor() const {
return factor_;
}
float_t scale() const {
return scale_;
}
private:
shape3d in_shape_;
float_t factor_;
float_t scale_;
};
} // namespace tiny_dnn
|
