gnn_tracking.metrics.binary_classification

This class collects figures of merit for binary classification

Module Contents

Classes

BinaryClassificationStats

Calculator for binary classification metrics.

Functions

zero_divide(→ float)	Normal division a/b but return 0 for x/0
get_maximized_bcs(→ dict[str, float])	Calculate the best possible binary classification stats for a given output and y.
roc_auc_score(→ float)	Wrapper that ignores exceptions
get_roc_auc_scores(true, predicted, max_fprs)	Calculate ROC AUC scores for a given set of maximum FPRs.

class gnn_tracking.metrics.binary_classification.BinaryClassificationStats(output: torch.Tensor, y: torch.Tensor, thld: torch.Tensor | float)

Calculator for binary classification metrics. All properties are cached, so they are only calculated once.

Parameters:

• output – Output weights

• y – True labels

• thld – Threshold to consider something true

Returns:

accuracy, TPR, TNR

_true() → torch.Tensor

n_true() → int

_false() → torch.Tensor

n_false() → int

_predicted_false() → torch.Tensor

n_predicted_false() → int

_predicted_true() → torch.Tensor

n_predicted_true() → int

TP() → float

TN() → float

FP() → float

FN() → float

acc() → float

TPR() → float

TNR() → float

FPR() → float

FNR() → float

balanced_acc() → float

F1() → float

MCC() → float

get_all() → dict[str, float]

gnn_tracking.metrics.binary_classification.zero_divide(a: float, b: float) → float

Normal division a/b but return 0 for x/0

gnn_tracking.metrics.binary_classification.get_maximized_bcs(*, output: torch.Tensor, y: torch.Tensor, n_samples=200) → dict[str, float]

Calculate the best possible binary classification stats for a given output and y.

Parameters:

• output – Weights

• y – True

• n_samples – Number of thresholds to sample

Returns:

Dictionary of metrics

gnn_tracking.metrics.binary_classification.roc_auc_score(*, y_true: torch.Tensor, y_score: torch.Tensor, max_fpr: float | None = None, device=None) → float

Wrapper that ignores exceptions that can e.g., be raised if there’s only one label present.

gnn_tracking.metrics.binary_classification.get_roc_auc_scores(true, predicted, max_fprs: Iterable[float | None])

Calculate ROC AUC scores for a given set of maximum FPRs.