libeblearn
Public Member Functions | Public Attributes | Protected Attributes
ebl::scalerclass_energy< T, Tstate > Class Template Reference
Inheritance diagram for ebl::scalerclass_energy< T, Tstate >:
ebl::l2_energy< T, Tstate > ebl::ebm_2< Tstate > ebl::module

List of all members.

Public Member Functions

 scalerclass_energy (bool apply_tanh=false, uint jsize=1, uint jitter_selection=0, float dist_coeff=1.0, float scale_coeff=1.0, bool predict_conf=false, bool predict_bconf=false, idx< T > *biases=NULL, idx< T > *coeffs=NULL, const char *name="scalerclass_energy")
virtual ~scalerclass_energy ()
 destructor.
virtual void fprop (Tstate &in, Tstate &scale, Tstate &energy)
 fprop: compute output from input
virtual void bprop (Tstate &in, Tstate &scale, Tstate &energy)
 bprop: compute gradient wrt inputs, given gradient wrt output
virtual void bbprop (Tstate &in, Tstate &scale, Tstate &energy)
 bprop: compute diaghession wrt inputs, given diaghessian wrt output
virtual void infer2 (Tstate &i1, Tstate &scale, infer_param &ip, Tstate *energy=NULL)
 compute value of in2 that minimizes the energy, given in1
virtual std::string describe ()
 Returns a string describing this module and its parameters.

Public Attributes

idx< T > last_target_raw
 Un-normalized target.
Tstate last_target
 Last target selected by fprop.

Protected Attributes

uint jsize
 Number of jitter elements.
bool apply_tanh
 If true, apply tanh to inputs.
uint jitter_selection
 Selection type.
float dist_coeff
 Coefficient of distance score.
float scale_coeff
 Coefficient of scale score.
Tstate tmp
 Temporary buffer.
Tstate tmp2
 Temporary buffer.
Tstate last_class_target
 Last class target selected.
Tstate last_jitt_target
 Last jitter target selected.
Tstate last_conf_target
 Last confidence target selected.
idx< T > best_target
 The last target selected by fprop.
tanh_module< T, Tstate > mtanh
 A tanh module.
bool predict_conf
 Predict confidence.
bool predict_bconf
 Predicted confidence is binary or not.
idx< T > * biases
 Normalization biases: scale,h,w,conf.
idx< T > * coeffs
 Normalization coeffs: scale,h,w,conf.

template<typename T, class Tstate = bbstate_idx<T>>
class ebl::scalerclass_energy< T, Tstate >

Constructor & Destructor Documentation

template<typename T , class Tstate >
ebl::scalerclass_energy< T, Tstate >::scalerclass_energy ( bool  apply_tanh = false,
uint  jsize = 1,
uint  jitter_selection = 0,
float  dist_coeff = 1.0,
float  scale_coeff = 1.0,
bool  predict_conf = false,
bool  predict_bconf = false,
idx< T > *  biases = NULL,
idx< T > *  coeffs = NULL,
const char *  name = "scalerclass_energy< T, Tstate >" 
)

constructor.

Parameters:
predict_confAdd an extra feature for confidence prediction.
predict_bconfIf true, predicted confidence is binary (0,1).
The documentation for this class was generated from the following files: