TarzaNN
TarzaNN neural network simulator
|
#include <GatingUnits.h>
Public Member Functions | |
GatingUnits (const int nXSize, const int nYSize, const int filterSizeX, const int filterSizeY, const int inputSizeX, const int inputSizeY) | |
~GatingUnits () | |
float | getGUMax () |
float | getGUMin () |
OutputImage * | stepGatingUnits (Matrix *gc, float max, float theta) |
OutputImage * | stepGatingUnitsLearn (Matrix *gc, float max, float theta, LearningFeaturePlane *fp) |
void | computeActivations (Matrix *activations) |
void | setZero () |
2D array of Matrix objects. Each Matrix object contains the gating units for the particular IU, for a given input
GatingUnits::GatingUnits | ( | const int | nXSize, |
const int | nYSize, | ||
const int | filterSizeX, | ||
const int | filterSizeY, | ||
const int | inputSizeX, | ||
const int | inputSizeY | ||
) |
GatingUnits::~GatingUnits | ( | ) |
Destructor. Deletes all the gating unit matrices, and the internal data structures
void GatingUnits::computeActivations | ( | Matrix * | activations | ) | [inline] |
float GatingUnits::getGUMax | ( | ) |
Get the maximum gating unit activation
float GatingUnits::getGUMin | ( | ) |
Get the minimum gating unit activation
void GatingUnits::setZero | ( | ) |
OutputImage * GatingUnits::stepGatingUnits | ( | Matrix * | gc, |
float | max, | ||
float | theta | ||
) |
Apply the STM WTA rule. For now, just threshold with theta, ignoring the temporal aspect of the competition
OutputImage * GatingUnits::stepGatingUnitsLearn | ( | Matrix * | gc, |
float | max, | ||
float | theta, | ||
LearningFeaturePlane * | fp | ||
) |