libeblearn
|
#include <ebl_cost.h>
Public Member Functions | |
logadd_layer (intg thick, intg si, intg sj) | |
void | fprop (fstate_idx< T > *in, fstate_idx< T > *out) |
void | bprop (fstate_idx< T > *in, fstate_idx< T > *out) |
void | bbprop (fstate_idx< T > *in, fstate_idx< T > *out) |
Public Attributes | |
idx< T > | expdist |
idx< T > | sumexp |
performs a log-add over spatial dimensions of an idx3-state output is an idx1-state
void ebl::logadd_layer< T >::bbprop | ( | fstate_idx< T > * | in, |
fstate_idx< T > * | out | ||
) |
this is not algebraically correct, but it's numerically more stable (at least, I think so).