Metrics¶
Aggregation Metrics¶
Computes Area Under the Curve (AUC) using the trapezoidal rule. 

Concatenate all input tensors along dimension dim. 

Calculate the maximum value of all elements in all the input tensors. 

Calculate the weighted mean value of all elements in all the input tensors. 

Calculate the minimum value of all elements in all the input tensors. 

Calculate the weighted sum value of all elements in all the input tensors. 

Calculate the throughput value which is the number of elements processed per second. 
Classification Metrics¶
Compute binary accuracy score, which is the frequency of input matching target. 

Compute AUPRC, also called Average Precision, which is the area under the PrecisionRecall Curve, for binary classification. 

Compute AUROC, which is the area under the ROC Curve, for binary classification. 

Compute AUROC, which is the area under the ROC Curve, for binary classification. 

Compute precision recall curve with given thresholds. 

Compute binary confusion matrix, a 2 by 2 tensor with counts ( (true positive, false negative) , (false positive, true negative) ) 

Compute binary f1 score, which is defined as the harmonic mean of precision and recall. 

Compute the normalized binary cross entropy between predicted input and groundtruth binary target. 

Compute the precision score for binary classification tasks, which is calculated as the ratio of the true positives and the sum of true positives and false positives. 

Returns precisionrecall pairs and their corresponding thresholds for binary classification tasks. 

Compute the recall score for binary classification tasks, which is calculated as the ratio of the true positives and the sum of true positives and false negatives. 

Returns the highest possible recall value give the minimum precision for binary classification tasks. 

Compute accuracy score, which is the frequency of input matching target. 

Compute AUPRC, also called Average Precision, which is the area under the PrecisionRecall Curve, for multiclass classification. 

Compute AUROC, which is the area under the ROC Curve, for multiclass classification in a one vs rest fashion. 

Compute AUROC, which is the area under the ROC Curve, for multiclass classification. 

Compute precision recall curve with given thresholds. 

Compute multiclass confusion matrix, a matrix of dimension num_classes x num_classes where each element at position (i,j) is the number of examples with true class i that were predicted to be class j. 

Compute f1 score, which is defined as the harmonic mean of precision and recall. 

Compute the precision score, the ratio of the true positives and the sum of true positives and false positives. 

Returns precisionrecall pairs and their corresponding thresholds for multiclass classification tasks. 

Compute the recall score, the ratio of the true positives and the sum of true positives and false negatives. 

Compute multilabel accuracy score, which is the frequency of input matching target. 

Compute AUPRC, also called Average Precision, which is the area under the PrecisionRecall Curve, for multilabel classification. 

Returns precisionrecall pairs and their corresponding thresholds for multilabel classification tasks. 

Returns the highest possible recall value given the minimum precision for each label and their corresponding thresholds for multilabel classification tasks. 

Compute multilabel accuracy score, which is the frequency of the top k label predicted matching target. 
Ranking Metrics¶
Compute the click through rate given click events. 

Compute the hit rate of the correct class among the top predicted classes. 

Compute the reciprocal rank of the correct class among the top predicted classes. 

Compute weighted calibration metric. 
Regression Metrics¶
Compute Mean Squared Error, which is the mean of squared error of input and target. 

Compute Rsquared score, which is the proportion of variance in the dependent variable that can be explained by the independent variable. 
Text Metrics¶
Compute BLEU score (https://en.wikipedia.org/wiki/BLEU) given translations and references. 

Perplexity measures how well a model predicts sample data. 

Compute the word error rate of the predicted word sequence(s) with the reference word sequence(s). 

Word Information Lost (WIL) is a metric of the performance of an automatic speech recognition system. 

Compute the word information preserved of the predicted word sequence(s) with the reference word sequence(s). 
Windowed Metrics¶
The windowed version of BinaryAUROC. 

The windowed version of BinaryNormalizedEntropy that provides both windowed and liftime values. 

The windowed version of ClickThroughRate that provides both windowed and lifetime values. 

The windowed version of Mean Squared Error that provides both windowed and liftime values. 

Compute weighted calibration metric. 