libeblearn
|
Creates a pyramid of the input with average subsampling. More...
#include <ebl_pooling.h>
Public Member Functions | |
average_pyramid_module (parameter< T, Tstate > *p, uint thickness, midxdim &strides, const char *name="average_pyramid_module") | |
virtual void | fprop (Tstate &in, Tstate &out) |
virtual void | fprop (Tstate &in, mstate< Tstate > &out) |
virtual void | bprop (Tstate &in, mstate< Tstate > &out) |
backward propagation from in to out (empty) | |
virtual void | bbprop (Tstate &in, mstate< Tstate > &out) |
bbackward propagation from in to out (empty) | |
virtual std::string | describe () |
Returns a string describing this module and its parameters. | |
virtual mfidxdim | bprop_size (mfidxdim &osize) |
Returns input dimensions corresponding to output dimensions 'osize'. | |
Protected Attributes | |
midxdim | strides |
Strides for each scale. | |
std::vector < subsampling_module< T, Tstate > * > | mods |
Subsampling modules. | |
bool | well_behaved |
True if strides increase by multiples of previous. |
Creates a pyramid of the input with average subsampling.
ebl::average_pyramid_module< T, Tstate >::average_pyramid_module | ( | parameter< T, Tstate > * | p, |
uint | thickness, | ||
midxdim & | strides, | ||
const char * | name = "average_pyramid_module< T, Tstate >" |
||
) |
Creates a pyramid with 'strides.size()' scales, each of which is constructed with an average subsampling with each stride.
p | is used to store all parametric variables in a single place. If p is null, a local buffer will be used. |
thickness | The number of features. |
void ebl::average_pyramid_module< T, Tstate >::fprop | ( | Tstate & | in, |
mstate< Tstate > & | out | ||
) | [virtual] |
Process 'in' into 'out' which will contain all scales separated in each state of the multi-state 'out'.
Reimplemented from ebl::s2m_module< T, Tstate >.