libeblearn
|
#include <ebl_energy.h>
Public Member Functions | |
cross_entropy_energy (const char *name="cross_entropy_energy") | |
virtual void | fprop (Tstate &in1, Tstate &in2, Tstate &energy) |
fprop: compute output from input | |
virtual void | bprop (Tstate &in1, Tstate &in2, Tstate &energy) |
bprop: compute gradient wrt inputs, given gradient wrt output | |
virtual void | bbprop (Tstate &in1, Tstate &in2, Tstate &energy) |
bprop: compute diaghession wrt inputs, given diaghessian wrt output | |
virtual void | infer2_copy (Tstate &i1, Tstate &i2, Tstate &energy) |
virtual std::string | describe () |
Returns a string describing this module and its parameters. | |
Protected Attributes | |
idx< T > | tmp |
Buffer for temporary operations. |
A cross-entropy energy module for multinomiol logistic regression. Energy equals: - desired where desired is a 1-of-n encoding of the groundtruth class.