Brush::Eval Namespace Reference

Classes
class	Evaluation
	Class for evaluating the fitness of individuals in a population. More...
class	Scorer
class	Scorer< P >

Functions
float	mse (const VectorXf &y, const VectorXf &yhat, VectorXf &loss, const vector< float > &class_weights)
	mean squared error
VectorXf	log_loss (const VectorXf &y, const VectorXf &predict_proba, const vector< float > &class_weights=vector< float >())
	Calculates the log loss between the predicted probabilities and the true labels.
float	mean_log_loss (const VectorXf &y, const VectorXf &predict_proba, VectorXf &loss, const vector< float > &class_weights)
	log loss
float	zero_one_loss (const VectorXf &y, const VectorXf &predict_proba, VectorXf &loss, const vector< float > &class_weights=vector< float >())
	Accuracy for binary classification.
float	bal_zero_one_loss (const VectorXf &y, const VectorXf &predict_proba, VectorXf &loss, const vector< float > &class_weights=vector< float >())
	Balanced accuracy for binary classification.
float	average_precision_score (const VectorXf &y, const VectorXf &predict_proba, VectorXf &loss, const vector< float > &class_weights=vector< float >())
	Calculates the average precision score between the predicted probabilities and the true labels.
float	precision_score (const VectorXf &y, const VectorXf &predict_proba, VectorXf &loss, const vector< float > &class_weights=vector< float >())
	Precision for binary classification (threshold 0.5, positive label 1).
float	recall_score (const VectorXf &y, const VectorXf &predict_proba, VectorXf &loss, const vector< float > &class_weights=vector< float >())
	Recall for binary classification (threshold 0.5, positive label 1).
float	roc_auc_score (const VectorXf &y, const VectorXf &predict_proba, VectorXf &loss, const vector< float > &class_weights=vector< float >())
	Area under the ROC curve for binary classification.
VectorXf	multi_log_loss (const VectorXf &y, const ArrayXXf &predict_proba, const vector< float > &class_weights=vector< float >())
	Calculates the multinomial log loss between the predicted probabilities and the true labels.
float	mean_multi_log_loss (const VectorXf &y, const ArrayXXf &predict_proba, VectorXf &loss, const vector< float > &class_weights=vector< float >())
	Calculates the mean multinomial log loss between the predicted probabilities and the true labels.
float	multi_zero_one_loss (const VectorXf &y, const ArrayXXf &predict_proba, VectorXf &loss, const vector< float > &class_weights=vector< float >())
	Accuracy for multi-classification.
float	multi_bal_zero_one_loss (const VectorXf &y, const ArrayXXf &predict_proba, VectorXf &loss, const vector< float > &class_weights=vector< float >())
	Balanced accuracy for multi-classification.
float	multi_precision_score (const VectorXf &y, const ArrayXXf &predict_proba, VectorXf &loss, const vector< float > &class_weights=vector< float >())
	Macro-averaged precision for multi-classification.
float	multi_recall_score (const VectorXf &y, const ArrayXXf &predict_proba, VectorXf &loss, const vector< float > &class_weights=vector< float >())
	Macro-averaged recall for multi-classification.
float	multi_roc_auc_score (const VectorXf &y, const ArrayXXf &predict_proba, VectorXf &loss, const vector< float > &class_weights=vector< float >())
	Macro-averaged one-vs-rest AUROC for multi-classification.
float	multi_average_precision_score (const VectorXf &y, const ArrayXXf &predict_proba, VectorXf &loss, const vector< float > &class_weights=vector< float >())
	Macro-averaged one-vs-rest average precision for multi-classification.

Function Documentation

average_precision_score()

float Brush::Eval::average_precision_score	(	const VectorXf &	y,
		const VectorXf &	predict_proba,
		VectorXf &	loss,
		const vector< float > &	class_weights = vector< float >() )

Calculates the average precision score between the predicted probabilities and the true labels.

Parameters

y	The true labels.
predict_proba	The predicted probabilities.
loss	Reference to store the calculated losses for each sample.
class_weights	The optional class weights.

Returns

The average precision score.

Definition at line 343 of file metrics.cpp.

Here is the call graph for this function:

Here is the caller graph for this function:

bal_zero_one_loss()

float Brush::Eval::bal_zero_one_loss	(	const VectorXf &	y,
		const VectorXf &	predict_proba,
		VectorXf &	loss,
		const vector< float > &	class_weights = vector< float >() )

Balanced accuracy for binary classification.

Parameters

y	The true labels.
predict_proba	The predicted probabilities.
loss	Reference to store the calculated losses for each sample.
class_weights	The optional class weights.

Returns

The final accuracy.

Definition at line 102 of file metrics.cpp.

Here is the call graph for this function:

Here is the caller graph for this function:

log_loss()

VectorXf Brush::Eval::log_loss	(	const VectorXf &	y,
		const VectorXf &	predict_proba,
		const vector< float > &	class_weights = vector< float >() )

Calculates the log loss between the predicted probabilities and the true labels.

Parameters

y	The true labels.
predict_proba	The predicted probabilities.
class_weights	The optional class weights.

Returns

The log loss.

Definition at line 19 of file metrics.cpp.

Here is the call graph for this function:

Here is the caller graph for this function:

mean_log_loss()

float Brush::Eval::mean_log_loss	(	const VectorXf &	y,
		const VectorXf &	predict_proba,
		VectorXf &	loss,
		const vector< float > &	class_weights = vector< float >() )

log loss

Calculates the mean log loss between the predicted probabilities and the true labels.

Parameters

y	The true labels.
predict_proba	The predicted probabilities.
loss	Reference to store the calculated losses for each sample.
class_weights	The optional class weights.

Returns

The mean log loss.

Definition at line 47 of file metrics.cpp.

Here is the call graph for this function:

Here is the caller graph for this function:

mean_multi_log_loss()

float Brush::Eval::mean_multi_log_loss	(	const VectorXf &	y,
		const ArrayXXf &	predict_proba,
		VectorXf &	loss,
		const vector< float > &	class_weights = vector< float >() )

Calculates the mean multinomial log loss between the predicted probabilities and the true labels.

Parameters

y	The true labels.
predict_proba	The predicted probabilities.
loss	Reference to store the calculated losses for each sample.
class_weights	The optional class weights.

Returns

The mean multinomial log loss.

Definition at line 417 of file metrics.cpp.

Here is the call graph for this function:

Here is the caller graph for this function:

mse()

float Brush::Eval::mse	(	const VectorXf &	y,
		const VectorXf &	yhat,
		VectorXf &	loss,
		const vector< float > &	class_weights = vector< float >() )

mean squared error

Calculates the mean squared error between the predicted values and the true values.

Parameters

y	The true values.
yhat	The predicted values.
loss	Reference to store the calculated losses for each sample.
class_weights	The optional class weights (not used for MSE).

Returns

The mean squared error.

Definition at line 11 of file metrics.cpp.

Here is the call graph for this function:

Here is the caller graph for this function:

multi_average_precision_score()

float Brush::Eval::multi_average_precision_score	(	const VectorXf &	y,
		const ArrayXXf &	predict_proba,
		VectorXf &	loss,
		const vector< float > &	class_weights = vector< float >() )

Macro-averaged one-vs-rest average precision for multi-classification.

Mean of the binary average precision of each class against the rest, skipping classes that are absent from y. The loss vector holds the per-sample multinomial log loss.

Definition at line 537 of file metrics.cpp.

Here is the call graph for this function:

Here is the caller graph for this function:

multi_bal_zero_one_loss()

float Brush::Eval::multi_bal_zero_one_loss	(	const VectorXf &	y,
		const ArrayXXf &	predict_proba,
		VectorXf &	loss,
		const vector< float > &	class_weights )

Balanced accuracy for multi-classification.

Definition at line 465 of file metrics.cpp.

Here is the call graph for this function:

Here is the caller graph for this function:

multi_log_loss()

VectorXf Brush::Eval::multi_log_loss	(	const VectorXf &	y,
		const ArrayXXf &	predict_proba,
		const vector< float > &	class_weights = vector< float >() )

Calculates the multinomial log loss between the predicted probabilities and the true labels.

Parameters

y	The true labels.
predict_proba	The predicted probabilities.
class_weights	The optional class weights.

Returns

The multinomial log loss.

Definition at line 396 of file metrics.cpp.

Here is the call graph for this function:

Here is the caller graph for this function:

multi_precision_score()

float Brush::Eval::multi_precision_score	(	const VectorXf &	y,
		const ArrayXXf &	predict_proba,
		VectorXf &	loss,
		const vector< float > &	class_weights = vector< float >() )

Macro-averaged precision for multi-classification.

Equivalent to sklearn's precision_score(average='macro', zero_division=0): averages over classes present in either the true or the predicted labels. The loss vector holds the misclassification indicator.

Definition at line 503 of file metrics.cpp.

Here is the call graph for this function:

Here is the caller graph for this function:

multi_recall_score()

float Brush::Eval::multi_recall_score	(	const VectorXf &	y,
		const ArrayXXf &	predict_proba,
		VectorXf &	loss,
		const vector< float > &	class_weights = vector< float >() )

Macro-averaged recall for multi-classification.

Equivalent to sklearn's recall_score(average='macro', zero_division=0). The loss vector holds the misclassification indicator.

Definition at line 509 of file metrics.cpp.

Here is the call graph for this function:

Here is the caller graph for this function:

multi_roc_auc_score()

float Brush::Eval::multi_roc_auc_score	(	const VectorXf &	y,
		const ArrayXXf &	predict_proba,
		VectorXf &	loss,
		const vector< float > &	class_weights = vector< float >() )

Macro-averaged one-vs-rest AUROC for multi-classification.

Mean of the binary AUROC of each class against the rest, skipping classes that are absent from y. The loss vector holds the per-sample multinomial log loss.

Definition at line 515 of file metrics.cpp.

Here is the call graph for this function:

Here is the caller graph for this function:

multi_zero_one_loss()

float Brush::Eval::multi_zero_one_loss	(	const VectorXf &	y,
		const ArrayXXf &	predict_proba,
		VectorXf &	loss,
		const vector< float > &	class_weights = vector< float >() )

Accuracy for multi-classification.

Parameters

y	The true labels.
predict_proba	The predicted probabilities.
loss	Reference to store the calculated losses for each sample.
class_weights	The optional class weights.

Returns

The average accuracy in a one-vs-all schema.

Definition at line 438 of file metrics.cpp.

Here is the call graph for this function:

Here is the caller graph for this function:

precision_score()

float Brush::Eval::precision_score	(	const VectorXf &	y,
		const VectorXf &	predict_proba,
		VectorXf &	loss,
		const vector< float > &	class_weights = vector< float >() )

Precision for binary classification (threshold 0.5, positive label 1).

Equivalent to sklearn's precision_score(zero_division=0). Class weights are used as sample weights. The loss vector holds the per-sample misclassification indicator (used in lexicase selection).

Parameters

y	The true labels.
predict_proba	The predicted probabilities.
loss	Reference to store the calculated losses for each sample.
class_weights	The optional class weights.

Returns

The precision.

Definition at line 360 of file metrics.cpp.

Here is the call graph for this function:

Here is the caller graph for this function:

recall_score()

float Brush::Eval::recall_score	(	const VectorXf &	y,
		const VectorXf &	predict_proba,
		VectorXf &	loss,
		const vector< float > &	class_weights = vector< float >() )

Recall for binary classification (threshold 0.5, positive label 1).

Equivalent to sklearn's recall_score(zero_division=0). The loss vector holds the per-sample misclassification indicator.

Parameters

y	The true labels.
predict_proba	The predicted probabilities.
loss	Reference to store the calculated losses for each sample.
class_weights	The optional class weights.

Returns

The recall.

Definition at line 374 of file metrics.cpp.

Here is the call graph for this function:

Here is the caller graph for this function:

roc_auc_score()

float Brush::Eval::roc_auc_score	(	const VectorXf &	y,
		const VectorXf &	predict_proba,
		VectorXf &	loss,
		const vector< float > &	class_weights = vector< float >() )

Area under the ROC curve for binary classification.

Equivalent to sklearn's roc_auc_score. Returns 0.5 when only one class is present (where the metric is undefined). The loss vector holds the per-sample log loss (used in lexicase selection).

Parameters

y	The true labels.
predict_proba	The predicted probabilities.
loss	Reference to store the calculated losses for each sample.
class_weights	The optional class weights.

Returns

The AUROC.

Definition at line 386 of file metrics.cpp.

Here is the call graph for this function:

Here is the caller graph for this function:

zero_one_loss()

float Brush::Eval::zero_one_loss	(	const VectorXf &	y,
		const VectorXf &	predict_proba,
		VectorXf &	loss,
		const vector< float > &	class_weights = vector< float >() )

Accuracy for binary classification.

Parameters

y	The true labels.
predict_proba	The predicted probabilities.
loss	Reference to store the calculated losses for each sample.
class_weights	The optional class weights.

Returns

The final accuracy.

Definition at line 75 of file metrics.cpp.

Here is the call graph for this function:

Here is the caller graph for this function: