File size: 10,076 Bytes
9375c9a |
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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 |
// Copyright (C) 2016 Davis E. King ([email protected])
// License: Boost Software License See LICENSE.txt for the full license.
#ifndef DLIB_DNn_UTILITIES_H_
#define DLIB_DNn_UTILITIES_H_
#include "core.h"
#include "utilities_abstract.h"
#include "../geometry.h"
#include <fstream>
namespace dlib
{
// ----------------------------------------------------------------------------------------
inline void randomize_parameters (
tensor& params,
unsigned long num_inputs_and_outputs,
dlib::rand& rnd
)
{
for (auto& val : params)
{
// Draw a random number to initialize the layer according to formula (16)
// from Understanding the difficulty of training deep feedforward neural
// networks by Xavier Glorot and Yoshua Bengio.
val = 2*rnd.get_random_float()-1;
val *= std::sqrt(6.0/(num_inputs_and_outputs));
}
}
// ----------------------------------------------------------------------------------------
namespace impl
{
class visitor_net_to_xml
{
public:
visitor_net_to_xml(std::ostream& out_) : out(out_) {}
template<typename input_layer_type>
void operator()(size_t idx, const input_layer_type& l)
{
out << "<layer idx='"<<idx<<"' type='input'>\n";
to_xml(l,out);
out << "</layer>\n";
}
template <typename T, typename U>
void operator()(size_t idx, const add_loss_layer<T,U>& l)
{
out << "<layer idx='"<<idx<<"' type='loss'>\n";
to_xml(l.loss_details(),out);
out << "</layer>\n";
}
template <typename T, typename U, typename E>
void operator()(size_t idx, const add_layer<T,U,E>& l)
{
out << "<layer idx='"<<idx<<"' type='comp'>\n";
to_xml(l.layer_details(),out);
out << "</layer>\n";
}
template <unsigned long ID, typename U, typename E>
void operator()(size_t idx, const add_tag_layer<ID,U,E>& /*l*/)
{
out << "<layer idx='"<<idx<<"' type='tag' id='"<<ID<<"'/>\n";
}
template <template<typename> class T, typename U>
void operator()(size_t idx, const add_skip_layer<T,U>& /*l*/)
{
out << "<layer idx='"<<idx<<"' type='skip' id='"<<(tag_id<T>::id)<<"'/>\n";
}
private:
std::ostream& out;
};
}
template <typename net_type>
void net_to_xml (
const net_type& net,
std::ostream& out
)
{
auto old_precision = out.precision(9);
out << "<net>\n";
visit_layers(net, impl::visitor_net_to_xml(out));
out << "</net>\n";
// restore the original stream precision.
out.precision(old_precision);
}
template <typename net_type>
void net_to_xml (
const net_type& net,
const std::string& filename
)
{
std::ofstream fout(filename);
net_to_xml(net, fout);
}
// ----------------------------------------------------------------------------------------
namespace impl
{
class visitor_net_map_input_to_output
{
public:
visitor_net_map_input_to_output(dpoint& p_) : p(p_) {}
dpoint& p;
template<typename input_layer_type>
void operator()(const input_layer_type& )
{
}
template <typename T, typename U>
void operator()(const add_loss_layer<T,U>& net)
{
(*this)(net.subnet());
}
template <typename T, typename U, typename E>
void operator()(const add_layer<T,U,E>& net)
{
(*this)(net.subnet());
p = net.layer_details().map_input_to_output(p);
}
template <bool B, typename T, typename U, typename E>
void operator()(const dimpl::subnet_wrapper<add_layer<T,U,E>,B>& net)
{
(*this)(net.subnet());
p = net.layer_details().map_input_to_output(p);
}
template <unsigned long ID, typename U, typename E>
void operator()(const add_tag_layer<ID,U,E>& net)
{
// tag layers are an identity transform, so do nothing
(*this)(net.subnet());
}
template <bool is_first, unsigned long ID, typename U, typename E>
void operator()(const dimpl::subnet_wrapper<add_tag_layer<ID,U,E>,is_first>& net)
{
// tag layers are an identity transform, so do nothing
(*this)(net.subnet());
}
template <template<typename> class TAG_TYPE, typename U>
void operator()(const add_skip_layer<TAG_TYPE,U>& net)
{
(*this)(layer<TAG_TYPE>(net));
}
template <bool is_first, template<typename> class TAG_TYPE, typename SUBNET>
void operator()(const dimpl::subnet_wrapper<add_skip_layer<TAG_TYPE,SUBNET>,is_first>& net)
{
// skip layers are an identity transform, so do nothing
(*this)(layer<TAG_TYPE>(net));
}
};
class visitor_net_map_output_to_input
{
public:
visitor_net_map_output_to_input(dpoint& p_) : p(p_) {}
dpoint& p;
template<typename input_layer_type>
void operator()(const input_layer_type& )
{
}
template <typename T, typename U>
void operator()(const add_loss_layer<T,U>& net)
{
(*this)(net.subnet());
}
template <typename T, typename U, typename E>
void operator()(const add_layer<T,U,E>& net)
{
p = net.layer_details().map_output_to_input(p);
(*this)(net.subnet());
}
template <bool B, typename T, typename U, typename E>
void operator()(const dimpl::subnet_wrapper<add_layer<T,U,E>,B>& net)
{
p = net.layer_details().map_output_to_input(p);
(*this)(net.subnet());
}
template <unsigned long ID, typename U, typename E>
void operator()(const add_tag_layer<ID,U,E>& net)
{
// tag layers are an identity transform, so do nothing
(*this)(net.subnet());
}
template <bool is_first, unsigned long ID, typename U, typename E>
void operator()(const dimpl::subnet_wrapper<add_tag_layer<ID,U,E>,is_first>& net)
{
// tag layers are an identity transform, so do nothing
(*this)(net.subnet());
}
template <template<typename> class TAG_TYPE, typename U>
void operator()(const add_skip_layer<TAG_TYPE,U>& net)
{
(*this)(layer<TAG_TYPE>(net));
}
template <bool is_first, template<typename> class TAG_TYPE, typename SUBNET>
void operator()(const dimpl::subnet_wrapper<add_skip_layer<TAG_TYPE,SUBNET>,is_first>& net)
{
// skip layers are an identity transform, so do nothing
(*this)(layer<TAG_TYPE>(net));
}
};
}
template <typename net_type>
inline dpoint input_tensor_to_output_tensor(
const net_type& net,
dpoint p
)
{
impl::visitor_net_map_input_to_output temp(p);
temp(net);
return p;
}
template <typename net_type>
inline dpoint output_tensor_to_input_tensor(
const net_type& net,
dpoint p
)
{
impl::visitor_net_map_output_to_input temp(p);
temp(net);
return p;
}
// ----------------------------------------------------------------------------------------
template <typename net_type>
size_t count_parameters(
const net_type& net
)
{
size_t num_parameters = 0;
visit_layer_parameters(net, [&](const tensor& t) { num_parameters += t.size(); });
return num_parameters;
}
// ----------------------------------------------------------------------------------------
namespace impl
{
class visitor_learning_rate_multiplier
{
public:
visitor_learning_rate_multiplier(double new_learning_rate_multiplier_) :
new_learning_rate_multiplier(new_learning_rate_multiplier_) {}
template <typename layer>
void operator()(layer& l) const
{
set_learning_rate_multiplier(l, new_learning_rate_multiplier);
}
private:
double new_learning_rate_multiplier;
};
}
template <typename net_type>
void set_all_learning_rate_multipliers(
net_type& net,
double learning_rate_multiplier
)
{
DLIB_CASSERT(learning_rate_multiplier >= 0);
impl::visitor_learning_rate_multiplier temp(learning_rate_multiplier);
visit_computational_layers(net, temp);
}
template <size_t begin, size_t end, typename net_type>
void set_learning_rate_multipliers_range(
net_type& net,
double learning_rate_multiplier
)
{
static_assert(begin <= end, "Invalid range");
static_assert(end <= net_type::num_layers, "Invalid range");
DLIB_CASSERT(learning_rate_multiplier >= 0);
impl::visitor_learning_rate_multiplier temp(learning_rate_multiplier);
visit_computational_layers_range<begin, end>(net, temp);
}
// ----------------------------------------------------------------------------------------
}
#endif // DLIB_DNn_UTILITIES_H_
|