class SpatialSubSampling

Class for doing sub-sampling over images.

Inheritance:

Public Fields

int k_w

Kernel size (height and width)

int d_x

'x' translation in the input image after each application of the kernel

int d_y

'y' translation in the input image after each application of the kernel

int n_input_planes

Number of input images.

int input_height

Height of each input image

int input_width

Width of each input image

int output_height

Height of each output image

int output_width

Width of each output image

real* weights

weights[i] means kernel-weight for output plane i.

real* der_weights

Derivatives associated to weights

real* biases

biases[i] is the bias for output plane i

real* der_biases

Derivatives associated to biases

Public Methods

SpatialSubSampling(int n_input_planes_, int width_, int height_, int k_w_=2, int d_x_=2, int d_y_=2)

Create a sub-sampling layer

Inherited from GradientMachine:

Public Fields

int n_inputs

int n_outputs

Parameters* params

Parameters* der_params

Sequence* beta

Public Methods

virtual void iterInitialize()

virtual void forward(Sequence* inputs)

virtual void backward(Sequence* inputs, Sequence* alpha)

virtual void setPartialBackprop(bool flag=true)

virtual void frameForward(int t, real* f_inputs, real* f_outputs)

virtual void frameBackward(int t, real* f_inputs, real* beta_, real* f_outputs, real* alpha_)

virtual void loadXFile(XFile* file)

virtual void saveXFile(XFile* file)

Inherited from Machine:

Public Fields

Sequence* outputs

Public Methods

virtual void reset()

virtual void setDataSet(DataSet* dataset_)

Inherited from Object:

Public Fields

Allocator* allocator

Public Methods

void addOption(const char* name, int size, void* ptr, const char* help="")

void addIOption(const char* name, int* ptr, int init_value, const char* help="")

void addROption(const char* name, real* ptr, real init_value, const char* help="")

void addBOption(const char* name, bool* ptr, bool init_value, const char* help="")

void addOOption(const char* name, Object** ptr, Object* init_value, const char* help="")

void setOption(const char* name, void* ptr)

void setIOption(const char* name, int option)

void setROption(const char* name, real option)

void setBOption(const char* name, bool option)

void setOOption(const char* name, Object* option)

void load(const char* filename)

void save(const char* filename)

void* operator new(size_t size, Allocator* allocator_=NULL)

void* operator new(size_t size, Allocator* allocator_, void* ptr_)

void operator delete(void* ptr)

Documentation

Class for doing sub-sampling over images.

Suppose you put n_input_planes images in each input frame. The images are in one big vector: each input frame has a size of n_input_planes*input_height*input_width. (image after image). Thus, n_inputs = n_input_planes*input_height*input_width.

Then, for each output planes, it takes its associated input plane and it computes the convolution of the input image with a kernel of size k_w*k_w, where the weights of the kernel are equals.

The output image size is computed in the constructor and put in output_height and output_width. n_outputs = n_input_planes*output_height*output_width.

Note that, depending of the size of your kernel, several (last) input columns or rows of the image could be lost.

Note also that no non-linearity is applied in this layer.

int k_w

Kernel size (height and width)

int d_x

'x' translation in the input image after each application of the kernel

int d_y

'y' translation in the input image after each application of the kernel

int n_input_planes

Number of input images. The number of output images in sub-sampling is the same.

int input_height

Height of each input image

int input_width

Width of each input image

int output_height

Height of each output image

int output_width

Width of each output image

real* weights

weights[i] means kernel-weight for output plane i. weights[i] contains only one weight.

real* der_weights

Derivatives associated to weights

real* biases

biases[i] is the bias for output plane i

real* der_biases

Derivatives associated to biases

SpatialSubSampling(int n_input_planes_, int width_, int height_, int k_w_=2, int d_x_=2, int d_y_=2)

Create a sub-sampling layer

This class has no child classes.

Author:

Ronan Collobert (collober@idiap.ch)

Alphabetic index HTML hierarchy of classes or Java