How well a classifier does on a labeled dataset.
More...
#include <performance.hpp>
|
| int64_t | n = 0 |
| |
| int64_t | classes = 0 |
| |
| double | accuracy = 0.0 |
| | The fraction of inputs whose most probable class is the label.
|
| |
| double | log_loss = 0.0 |
| | The mean negative log probability of the label (scikit-learn's log_loss). A probability of 0 counts as 1e-15.
|
| |
| double | brier = 0.0 |
| | Brier's multi-class score: the mean over inputs of sum_k (p_k - y_k)^2, from 0 to 2 (scikit-learn's brier_score_loss with scale_by_half=False).
|
| |
| double | ece = 0.0 |
| | The expected calibration error: sum over bins of count/n |accuracy - confidence|.
|
| |
| double | mce = 0.0 |
| | The maximum calibration error: the largest |accuracy - confidence| of a non-empty bin.
|
| |
| std::vector< ReliabilityBin > | bins |
| | The reliability diagram: one entry per bin, from confidence 0 to 1.
|
| |
| std::vector< std::vector< int64_t > > | confusion |
| | confusion[true class][predicted class]: how many inputs.
|
| |
| std::vector< double > | precision |
| | Per class: the fraction of its predictions that are right, of its inputs found, and their harmonic mean; 0 where there is nothing to divide by, as scikit-learn's zero_division=0.
|
| |
| std::vector< double > | recall |
| |
| std::vector< double > | f1 |
| |
| std::vector< int64_t > | support |
| | Per class: how many inputs have it as their label.
|
| |
| double | macro_f1 = 0.0 |
| | The unweighted mean of f1 over the classes.
|
| |
How well a classifier does on a labeled dataset.
◆ accuracy
| double pulsatrix::ClassificationReport::accuracy = 0.0 |
The fraction of inputs whose most probable class is the label.
◆ bins
The reliability diagram: one entry per bin, from confidence 0 to 1.
◆ brier
| double pulsatrix::ClassificationReport::brier = 0.0 |
Brier's multi-class score: the mean over inputs of sum_k (p_k - y_k)^2, from 0 to 2 (scikit-learn's brier_score_loss with scale_by_half=False).
◆ classes
| int64_t pulsatrix::ClassificationReport::classes = 0 |
◆ confusion
| std::vector<std::vector<int64_t> > pulsatrix::ClassificationReport::confusion |
confusion[true class][predicted class]: how many inputs.
◆ ece
| double pulsatrix::ClassificationReport::ece = 0.0 |
The expected calibration error: sum over bins of count/n |accuracy - confidence|.
◆ f1
| std::vector<double> pulsatrix::ClassificationReport::f1 |
◆ log_loss
| double pulsatrix::ClassificationReport::log_loss = 0.0 |
The mean negative log probability of the label (scikit-learn's log_loss). A probability of 0 counts as 1e-15.
◆ macro_f1
| double pulsatrix::ClassificationReport::macro_f1 = 0.0 |
The unweighted mean of f1 over the classes.
◆ mce
| double pulsatrix::ClassificationReport::mce = 0.0 |
The maximum calibration error: the largest |accuracy - confidence| of a non-empty bin.
| int64_t pulsatrix::ClassificationReport::n = 0 |
◆ precision
| std::vector<double> pulsatrix::ClassificationReport::precision |
Per class: the fraction of its predictions that are right, of its inputs found, and their harmonic mean; 0 where there is nothing to divide by, as scikit-learn's zero_division=0.
◆ recall
| std::vector<double> pulsatrix::ClassificationReport::recall |
◆ support
| std::vector<int64_t> pulsatrix::ClassificationReport::support |
Per class: how many inputs have it as their label.
The documentation for this struct was generated from the following file: