-
Notifications
You must be signed in to change notification settings - Fork 7
Expand file tree
/
Copy pathCNN.h
More file actions
204 lines (122 loc) · 4.22 KB
/
CNN.h
File metadata and controls
204 lines (122 loc) · 4.22 KB
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
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
#ifndef CNN_H
#define CNN_H
#include "Matrix.h"
#include "Layer.h"
#include "lInput.h"
#include "lConv.h"
#include "lPool.h"
#include "lFlatten.h"
#include "lDense.h"
#include "DataHandler.h"
class CNN {
private:
int feature_maps;
int in_rows;
int in_cols;
int out_rows;
int out_cols;
int nlayers;
public:
std::vector<Layer*> layers;
//Constructor
CNN( int in_rows, int in_cols, int feature_maps ) {
//Set dimensions
this->in_rows = in_rows;
this->in_cols = in_cols;
this->feature_maps = feature_maps;
out_rows = in_rows;
out_cols = in_cols;
//Add input layer
layers.push_back(new lInput(feature_maps, in_rows, in_cols));
nlayers = 1;
}
//Properties
int getLayers() { return nlayers; }
int getOutRows() { return out_rows; }
int getOutCols() { return out_cols; }
int getFeatureMaps() { return feature_maps; }
//Functions
void addConv(int k_size) {
layers.push_back(new lConv(feature_maps, layers[nlayers-1]->out_rows, layers[nlayers-1]->out_cols, k_size));
nlayers++;
out_rows = layers[nlayers- 1]->out_rows;
out_cols = layers[nlayers - 1]->out_cols;
return;
}
void addPool(int p_size) {
layers.push_back(new lPool(feature_maps, layers[nlayers-1]->out_rows, layers[nlayers-1]->out_cols, p_size));
nlayers++;
out_rows = layers[nlayers- 1]->out_rows;
out_cols = layers[nlayers - 1]->out_cols;
return;
}
void addFlatten() {
layers.push_back(new lFlatten(feature_maps, layers[nlayers-1]->out_rows, layers[nlayers-1]->out_cols));
nlayers++;
out_rows = layers[nlayers- 1]->out_rows;
out_cols = layers[nlayers - 1]->out_cols;
return;
}
void addDense(int out_size) {
layers.push_back(new lDense(layers[nlayers-1]->out_rows, out_size));
nlayers++;
out_rows = layers[nlayers- 1]->out_rows;
out_cols = layers[nlayers - 1]->out_cols;
return;
}
Tensor feedforward( Tensor in ) {
for (int i = 0; i < nlayers; i++) {
layers[i]->feedforward(in);
in = layers[i]->out.copy();
}
return layers[nlayers-1]->out;
}
double train( Tensor in, Tensor target, double lrate ) {
//Feedforward inputs
feedforward(in);
//Get deltas at output and cost
Tensor delta(1, out_rows, out_cols);
double cost = getCost(target, delta);
for (int i = nlayers - 1; i > 0; i--) {
delta = layers[i]->feedback(delta).copy();
//Update weights
layers[i]->updateweights( lrate );
}
return cost;
}
double getCost( Tensor target ) {
double cost = 0.0;
for (int i = 0; i < out_rows; i++) {
for (int j = 0; j < out_cols; j++) {
cost += pow(layers[nlayers-1]->out(0, i, j) - target(0, i, j), 2);
}
}
cost /= (2 * out_rows * out_cols);
return cost;
}
double getCost( Tensor target, Tensor & delta ) {
double cost = 0.0;
for (int i = 0; i < out_rows; i++) {
for (int j = 0; j < out_cols; j++) {
delta(0, i, j) = (layers[nlayers - 1]->out(0, i, j) - target(0, i, j)) / (out_rows * out_cols);
cost += pow(layers[nlayers - 1]->out(0, i, j) - target(0, i, j), 2);
}
}
cost /= (2 * out_rows * out_cols);
return cost;
}
void printWeights( int layer_num ) {
if (layer_num > 0 && layer_num < nlayers)
layers[layer_num]->getWeights().print();
return;
}
void print() {
std::stringstream s;
s << "[Network] In: " << in_rows << "x" << in_cols << " | Out: " << out_rows << "x" << out_cols << " | Feature Maps: " << feature_maps << std::endl;
for (int i = 0; i < nlayers; i ++)
s << "[Layer " << (i + 1) << "]: Type: " << layers[i]->getType() << " | Out: " << layers[i]->getDim() << "x" << layers[i]->getRows() << "x" << layers[i]->getCols() << std::endl;
std::cout << s.str();
return;
}
};
#endif