mlnext.score

Module for model evaluation.

Functions

apply_threshold	Applies threshold t to x.
auc_point_adjust_metrics	Calculates the area under the curve for performance metrics with point-adjusted predictions for values of k in [0,100].
bern_log_likelihood	Calculates the log likelihood of x being produced by a bernoulli distribution parameterized by mean.
eval_metrics	Calculates accuracy, f1, precision, recall and recall_anomalies.
eval_metrics_all	Calculates combined accuracy, f1, precision, recall and AUC scores for multiple arrays.
eval_sigmoid	Turns a binary-class sigmoid prediction into 0-1 class labels.
eval_softmax	Turns a multi-class softmax prediction into class labels.
get_threshold	Returns the perc-th percentile of x.
kl_divergence	Calculates the kl divergence kld(q||p) between a normal gaussian p (prior_mean, prior_std) and a normal distribution q parameterized by mean and log_var.
l2_norm	Calculates the l2-norm (euclidean distance) for x and x_hat.
moving_average	Calculates the moving average for X with stepsize step.
norm_log_likelihood	Calculates the negative log likelihood that x was drawn from a normal gaussian distribution defined by mean and log_var.
point_adjust_metrics	Calculates the performance metrics for various k in [0, 100].
pr_curve	Computes precision-recall pairs for different probability thresholds for binary classification tasks.

Classes

ConfusionMatrix	ConfusionMatrix is a confusion matrix for a binary classification problem.
PRCurve	Container for the result of pr_curve.