|
libeblearn
|
#include <ebl_answer.h>
Public Member Functions | |
| trainable_module (ebm_2< Tin1, Tin2, Ten > &energy, module_1_1< T, Tin1 > &mod1, module_1_1< T, Tin2 > *mod2=NULL, answer_module< T, Tds1, Tds2, Tin1 > *dsmod1=NULL, answer_module< T, Tds1, Tds2, Tin2 > *dsmod2=NULL, const char *name="trainable_module", const char *switcher="") | |
| virtual void | fprop (labeled_datasource< T, Tds1, Tds2 > &ds, Ten &energy) |
| virtual void | bprop (labeled_datasource< T, Tds1, Tds2 > &ds, Ten &energy) |
| virtual void | bbprop (labeled_datasource< T, Tds1, Tds2 > &ds, Ten &energy) |
| virtual int | infer2 (labeled_datasource< T, Tds1, Tds2 > &ds, Ten &energy) |
| virtual void | forget (forget_param_linear &fp) |
| Randomize internal modules using randomization parameters 'fp'. | |
| virtual const Tin1 & | compute_answers () |
| virtual void | compute_answers (Tin1 &ans) |
| virtual bool | correct (Tin1 &answer, Tin1 &label) |
| Returns true if 'answer' matches with 'label'. | |
| virtual void | update_log (classifier_meter &log, intg age, idx< T > &energy, idx< T > &answer, idx< T > &label, idx< T > &target, idx< T > &rawout) |
| Update the 'log' according to the type of answer module. | |
| virtual idx< T > | compute_targets (labeled_datasource< T, Tds1, Tds2 > &ds) |
| virtual const char * | name () |
| Returns the name of this module. | |
| virtual std::string | describe () |
| Returns a string describing this module and its parameters. | |
Protected Member Functions | |
| void | update_scale (labeled_datasource< T, Tds1, Tds2 > &ds) |
| Update the scale switch if an ms module was found in mod1. | |
Protected Attributes | |
| ebm_2< Tin1, Tin2, Ten > & | energy_mod |
| module_1_1< T, Tin1, Tin1 > & | mod1 |
| module_1_1< T, Tin1, Tin1 > * | mod2 |
|
answer_module< T, Tds1, Tds2, Tin1 > * | dsmod1 |
|
answer_module< T, Tds1, Tds2, Tin2 > * | dsmod2 |
| mstate< Tin1 > | msin1 |
| Tin1 | in1 |
| Tin1 | out1 |
| Output of mod1. | |
| Tin2 | in2 |
| Tin2 | out2 |
| Tin1 | answers |
| Tin1 | targets |
| string | mod_name |
| Ten | tmp_energy |
| ms_module< T, Tin1 > * | ms_switch |
Friends | |
| class | trainable_module_gui |
| class | supervised_trainer |
| class | supervised_trainer_gui |
A module that takes a datasource 'ds' as input and produces an energy as output. Inputs will be gathered from the datasource and fed into a 2-input energy module. The way the inputs are gathered from the datasource can optionally be defined by specific information collection modules.
| ebl::trainable_module< T, Tds1, Tds2, Tin1, Tin2, Ten >::trainable_module | ( | ebm_2< Tin1, Tin2, Ten > & | energy, |
| module_1_1< T, Tin1 > & | mod1, | ||
| module_1_1< T, Tin2 > * | mod2 = NULL, |
||
| answer_module< T, Tds1, Tds2, Tin1 > * | dsmod1 = NULL, |
||
| answer_module< T, Tds1, Tds2, Tin2 > * | dsmod2 = NULL, |
||
| const char * | name = "trainable_module< T, Tds1, Tds2, Tin1, Tin2, Ten >", |
||
| const char * | switcher = "" |
||
| ) |
Constructor for ebm_2, i.e. 2 inputs producing an energy.
| energy | The energy module. |
| mod1 | The 1st module. |
| mod2 | The 2nd module, optional because sometimes no processing is required on the 2nd flow of data. |
| dsmod1 | The module that picks/preprocesses data from ds to mod1, by default (null), will simply pass 'input' data from ds. |
| dsmod2 | The module that picks/preprocesses data from ds to mod2, by default (null), will simply pass 'label' data from ds. |
| switcher | Name of the module used to switch based on scale info. |
| const Tin1 & ebl::trainable_module< T, Tds1, Tds2, Tin1, Tin2, Ten >::compute_answers | ( | ) | [virtual] |
Compute answers corresponding to the last input passed through fprop() and return a reference to the answer matrix. The answer matrix depends on the answer module defined as 'dsmod2'.
| void ebl::trainable_module< T, Tds1, Tds2, Tin1, Tin2, Ten >::compute_answers | ( | Tin1 & | ans | ) | [virtual] |
Compute answers corresponding to the last input passed through fprop() and put them into 'ans'. The answer matrix depends on the answer module defined as 'dsmod2'.
| idx< T > ebl::trainable_module< T, Tds1, Tds2, Tin1, Tin2, Ten >::compute_targets | ( | labeled_datasource< T, Tds1, Tds2 > & | ds | ) | [virtual] |
Compute targets corresponding to the current sample of ds. and return a reference to the answer matrix. The targets matrix depends on the answer module defined as 'dsmod2'.