mlpack 3.4.2
mean_pooling.hpp
Go to the documentation of this file.
1
13#ifndef MLPACK_METHODS_ANN_LAYER_MEAN_POOLING_HPP
14#define MLPACK_METHODS_ANN_LAYER_MEAN_POOLING_HPP
15
16#include <mlpack/prereqs.hpp>
17
18namespace mlpack {
19namespace ann {
20
29template <
30 typename InputDataType = arma::mat,
31 typename OutputDataType = arma::mat
32>
34{
35 public:
38
48 MeanPooling(const size_t kernelWidth,
49 const size_t kernelHeight,
50 const size_t strideWidth = 1,
51 const size_t strideHeight = 1,
52 const bool floor = true);
53
61 template<typename eT>
62 void Forward(const arma::Mat<eT>& input, arma::Mat<eT>& output);
63
73 template<typename eT>
74 void Backward(const arma::Mat<eT>& /* input */,
75 const arma::Mat<eT>& gy,
76 arma::Mat<eT>& g);
77
79 OutputDataType const& OutputParameter() const { return outputParameter; }
81 OutputDataType& OutputParameter() { return outputParameter; }
82
84 OutputDataType const& Delta() const { return delta; }
86 OutputDataType& Delta() { return delta; }
87
89 size_t const& InputWidth() const { return inputWidth; }
91 size_t& InputWidth() { return inputWidth; }
92
94 size_t const& InputHeight() const { return inputHeight; }
96 size_t& InputHeight() { return inputHeight; }
97
99 size_t const& OutputWidth() const { return outputWidth; }
101 size_t& OutputWidth() { return outputWidth; }
102
104 size_t const& OutputHeight() const { return outputHeight; }
106 size_t& OutputHeight() { return outputHeight; }
107
109 size_t InputSize() const { return inSize; }
110
112 size_t OutputSize() const { return outSize; }
113
115 size_t KernelWidth() const { return kernelWidth; }
117 size_t& KernelWidth() { return kernelWidth; }
118
120 size_t KernelHeight() const { return kernelHeight; }
122 size_t& KernelHeight() { return kernelHeight; }
123
125 size_t StrideWidth() const { return strideWidth; }
127 size_t& StrideWidth() { return strideWidth; }
128
130 size_t StrideHeight() const { return strideHeight; }
132 size_t& StrideHeight() { return strideHeight; }
133
135 bool const& Floor() const { return floor; }
137 bool& Floor() { return floor; }
138
140 bool Deterministic() const { return deterministic; }
142 bool& Deterministic() { return deterministic; }
143
147 template<typename Archive>
148 void serialize(Archive& ar, const unsigned int /* version */);
149
150 private:
157 template<typename eT>
158 void Pooling(const arma::Mat<eT>& input, arma::Mat<eT>& output)
159 {
160 for (size_t j = 0, colidx = 0; j < output.n_cols;
161 ++j, colidx += strideHeight)
162 {
163 for (size_t i = 0, rowidx = 0; i < output.n_rows;
164 ++i, rowidx += strideWidth)
165 {
166 arma::mat subInput = input(
167 arma::span(rowidx, rowidx + kernelWidth - 1 - offset),
168 arma::span(colidx, colidx + kernelHeight - 1 - offset));
169
170 output(i, j) = arma::mean(arma::mean(subInput));
171 }
172 }
173 }
174
181 template<typename eT>
182 void Unpooling(const arma::Mat<eT>& input,
183 const arma::Mat<eT>& error,
184 arma::Mat<eT>& output)
185 {
186 const size_t rStep = input.n_rows / error.n_rows - offset;
187 const size_t cStep = input.n_cols / error.n_cols - offset;
188
189 arma::Mat<eT> unpooledError;
190 for (size_t j = 0; j < input.n_cols - cStep; j += cStep)
191 {
192 for (size_t i = 0; i < input.n_rows - rStep; i += rStep)
193 {
194 const arma::Mat<eT>& inputArea = input(arma::span(i, i + rStep - 1),
195 arma::span(j, j + cStep - 1));
196
197 unpooledError = arma::Mat<eT>(inputArea.n_rows, inputArea.n_cols);
198 unpooledError.fill(error(i / rStep, j / cStep) / inputArea.n_elem);
199
200 output(arma::span(i, i + rStep - 1 - offset),
201 arma::span(j, j + cStep - 1 - offset)) += unpooledError;
202 }
203 }
204 }
205
207 size_t kernelWidth;
208
210 size_t kernelHeight;
211
213 size_t strideWidth;
214
216 size_t strideHeight;
217
219 bool floor;
220
222 size_t inSize;
223
225 size_t outSize;
226
228 size_t inputWidth;
229
231 size_t inputHeight;
232
234 size_t outputWidth;
235
237 size_t outputHeight;
238
240 bool reset;
241
243 bool deterministic;
244
246 size_t offset;
247
249 size_t batchSize;
250
252 arma::cube outputTemp;
253
255 arma::cube inputTemp;
256
258 arma::cube gTemp;
259
261 OutputDataType delta;
262
264 OutputDataType gradient;
265
267 OutputDataType outputParameter;
268}; // class MeanPooling
269
270
271} // namespace ann
272} // namespace mlpack
273
274// Include implementation.
275#include "mean_pooling_impl.hpp"
276
277#endif
Implementation of the MeanPooling.
OutputDataType const & Delta() const
Get the delta.
size_t StrideHeight() const
Get the stride height.
size_t const & OutputHeight() const
Get the output height.
size_t & InputHeight()
Modify the input height.
bool const & Floor() const
Get the value of the rounding operation.
void Forward(const arma::Mat< eT > &input, arma::Mat< eT > &output)
Ordinary feed forward pass of a neural network, evaluating the function f(x) by propagating the activ...
OutputDataType const & OutputParameter() const
Get the output parameter.
MeanPooling()
Create the MeanPooling object.
size_t InputSize() const
Get the input size.
size_t & InputWidth()
Modify the input width.
size_t KernelWidth() const
Get the kernel width.
MeanPooling(const size_t kernelWidth, const size_t kernelHeight, const size_t strideWidth=1, const size_t strideHeight=1, const bool floor=true)
Create the MeanPooling object using the specified number of units.
size_t KernelHeight() const
Get the kernel height.
size_t & StrideWidth()
Modify the stride width.
size_t & KernelWidth()
Modify the kernel width.
size_t & OutputHeight()
Modify the output height.
size_t OutputSize() const
Get the output size.
bool & Deterministic()
Modify the value of the deterministic parameter.
bool Deterministic() const
Get the value of the deterministic parameter.
size_t const & InputWidth() const
Get the intput width.
bool & Floor()
Modify the value of the rounding operation.
size_t & KernelHeight()
Modify the kernel height.
size_t StrideWidth() const
Get the stride width.
size_t & StrideHeight()
Modify the stride height.
void Backward(const arma::Mat< eT > &, const arma::Mat< eT > &gy, arma::Mat< eT > &g)
Ordinary feed backward pass of a neural network, using 3rd-order tensors as input,...
OutputDataType & OutputParameter()
Modify the output parameter.
size_t & OutputWidth()
Modify the output width.
size_t const & InputHeight() const
Get the input height.
size_t const & OutputWidth() const
Get the output width.
void serialize(Archive &ar, const unsigned int)
Serialize the layer.
OutputDataType & Delta()
Modify the delta.
Linear algebra utility functions, generally performed on matrices or vectors.
Definition: cv.hpp:1
The core includes that mlpack expects; standard C++ includes and Armadillo.