libeblearn
ebl::convabsnorm_layer< T, Tstate > Class Template Reference

#include <ebl_layers.h>

Inheritance diagram for ebl::convabsnorm_layer< T, Tstate >:
ebl::module_1_1< T, Tstate > ebl::module

List of all members.

Public Member Functions

 convabsnorm_layer (parameter< T, Tstate > *p, intg kerneli, intg kernelj, intg stridei, intg stridej, idx< intg > &tbl, bool mirror=false, bool tanh=true, const char *name="convabsnorm_layer")
virtual ~convabsnorm_layer ()
 Destructor.
void fprop (Tstate &in, Tstate &out)
 fprop from in to out
void bprop (Tstate &in, Tstate &out)
 bprop
void bbprop (Tstate &in, Tstate &out)
 bbprop
void forget (forget_param_linear &fp)
 initialize the weights to random values
virtual fidxdim fprop_size (fidxdim &i_size)
virtual fidxdim bprop_size (const fidxdim &o_size)
virtual convabsnorm_layer< T,
Tstate > * 
copy ()
 Returns a deep copy of this module.

Public Attributes

convolution_layer< T, Tstate > lconv
 convolution layer
abs_module< T, Tstate > abs
 absolute rectification
contrast_norm_module< T, Tstate > norm
 constrast normalization
Tstate * tmp
 temporary results
Tstate * tmp2
 temporary results

Detailed Description

template<typename T, class Tstate = bbstate_idx<T>>
class ebl::convabsnorm_layer< T, Tstate >

a convolution layer with absolute rectification and constrast normalization


Constructor & Destructor Documentation

template<typename T , class Tstate >
ebl::convabsnorm_layer< T, Tstate >::convabsnorm_layer ( parameter< T, Tstate > *  p,
intg  kerneli,
intg  kernelj,
intg  stridei,
intg  stridej,
idx< intg > &  tbl,
bool  mirror = false,
bool  tanh = true,
const char *  name = "convabsnorm_layer< T, Tstate >" 
)

constructor. Arguments are a pointer to a parameter in which the trainable weights will be appended, the number of inputs, and the number of outputs.

Parameters:
pis used to store all parametric variables in a single place.
kerneliis the height of the convolution kernel
kerneljis the width of the convolution kernel
strideiis the stride at which convolutions are done on the height axis.
stridejis the stride at which convolutions are done on the width axis.
tableis the convolution connection table.
tanhIf true, use tanh squasher, stdsigmoid otherwise.

Member Function Documentation

template<typename T , class Tstate >
fidxdim ebl::convabsnorm_layer< T, Tstate >::bprop_size ( const fidxdim &  o_size) [virtual]

Return dimensions compatible with this module given output dimensions. See module_1_1_gen's documentation for more details.

Reimplemented from ebl::module_1_1< T, Tstate >.

template<typename T , class Tstate >
fidxdim ebl::convabsnorm_layer< T, Tstate >::fprop_size ( fidxdim &  i_size) [virtual]

Return dimensions that are compatible with this module. See module_1_1_gen's documentation for more details.

Reimplemented from ebl::module_1_1< T, Tstate >.


The documentation for this class was generated from the following files: