libeblearn
ebl::fc_ebm2< T, Tin1, Tin2, Ten > Class Template Reference

#include <ebl_arch.h>

Inheritance diagram for ebl::fc_ebm2< T, Tin1, Tin2, Ten >:
ebl::ebm_2< Tin1, Tin2, Ten > ebl::module

List of all members.

Public Member Functions

 fc_ebm2 (module_1_1< T, Tin1 > &fm, Tin1 &fo, ebm_2< Tin1, Tin2, Ten > &fc)
virtual void fprop (Tin1 &in1, Tin2 &in2, Ten &energy)
 fprop: compute output from input
virtual void bprop (Tin1 &in1, Tin2 &in2, Ten &energy)
 bprop: compute gradient wrt inputs, given gradient wrt output
virtual void bbprop (Tin1 &in1, Tin2 &in2, Ten &energy)
 bprop: compute diaghession wrt inputs, given diaghessian wrt output
virtual void forget (forget_param_linear &fp)
virtual double infer2 (Tin1 &i1, Tin2 &i2, infer_param &ip, Tin2 *label=NULL, Ten *energy=NULL)

Public Attributes

module_1_1< T, Tin1, Tin1 > & fmod
Tin1 & fout
ebm_2< Tin1, Tin2, Ten > & fcost

Detailed Description

template<typename T, class Tin1 = bbstate_idx<T>, class Tin2 = Tin1, class Ten = Tin1>
class ebl::fc_ebm2< T, Tin1, Tin2, Ten >

standard 2 input EBM with one module-1-1, and one ebm-2 on top. fc stands for "function+cost".


Member Function Documentation

template<typename T , class Tin1, class Tin2, class Ten>
double ebl::fc_ebm2< T, Tin1, Tin2, Ten >::infer2 ( Tin1 &  i1,
Tin2 &  i2,
infer_param ip,
Tin2 *  label = NULL,
Ten *  energy = NULL 
) [virtual]

compute value of in2 that minimizes the energy, given in1 if label is given, fill the corresponding energy.

Reimplemented from ebl::ebm_2< Tin1, Tin2, Ten >.


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