AdaBoost W+ metric (or any metric based on weighted pattern sum) for DecisionStumpPolicy (in this case only used as container for the param recovered) using PatternResponse.
More...
#include <WeightedPatternResponse.h>
|
| static double | optimize (DecisionStump< int > &h, const PatternResponse *store, int n, double sum_d) __attribute__((deprecated)) |
| |
| static double | optimize (DecisionStump< int > &h, const PatternResponse *store, int n, double wp, double wn, bool strictly_growing=false) |
| |
| static double | optimize (MultiClassDecisionStumpSingleThreshold< int > &h, const WeightedPatternResponse< int > *store, int n, const double *sum_d) |
| |
| static double | optimize (MultiClassDecisionStumpMultiThreshold< int > &h, const WeightedPatternResponse< int > *store, int n, const double *sum_d) |
| |
AdaBoost W+ metric (or any metric based on weighted pattern sum) for DecisionStumpPolicy (in this case only used as container for the param recovered) using PatternResponse.
This function find the best threshold in a DecisionStump maximing W+ parameters (elements over threshold weighted negatively, element lesser or equal of threshold weighted positively)
- Parameters
-
| h | a DecisionStumpPolicy (in output the threshold and parity fields are setted) |
| store | an array of (sorted) value returned by HaarClassifier |
| n | number of stored elements |
| sum_d | sum of store.d element with sign (precomputed) |
- Returns
- the adaboost W+ parameters and set h threshold TODO: change name
| static double AdaBoostMetric::optimize |
( |
DecisionStump< int > & |
h, |
|
|
const PatternResponse * |
store, |
|
|
int |
n, |
|
|
double |
wp, |
|
|
double |
wn, |
|
|
bool |
strictly_growing = false |
|
) |
| |
|
static |
This function find the best threshold in a DecisionStump maximing W+ parameters (elements over threshold weighted negatively, element lesser or equal of threshold weighted positively)
- Parameters
-
| h | a DecisionStumpPolicy (in output the threshold and parity fields are setted) |
| store | an array of (sorted) value returned by HaarClassifier |
| n | number of stored elements |
| wp | precomputed sum of positive elements |
| wn | precomputed sum of negative elements |
| strictly_growing | suggest that elements are strictly growing, in order to avoid check element transitions |
- Returns
- the adaboost W+ parameters and set h threshold
optimize the multiclass problem, single threshold TODO: unimplemented
optimize the multiclass problem, multiple threshold TODO: unimplemented
The documentation for this class was generated from the following file:
