Update mc_auroc.py

mc_auroc.py

@@ -85,7 +85,7 @@ class AVG_MULTICLASS_AUROC(evaluate.Metric):
 
         return lower, upper
 
-    def _compute(self, predictions: Sequence[Sequence[float]], references: Sequence[int]):
+    def _compute(self, predictions: Sequence[Sequence[float]], references: Sequence[int], CI=False):
         """
         Computes the average AUROC score for multi-class classification problems.
         """
@@ -102,17 +102,17 @@ class AVG_MULTICLASS_AUROC(evaluate.Metric):
             fpr[i], tpr[i], thresholds[i] = roc_curve(y_true=[1 if x == n_classes[i] else 0 for x in references],
                                                       y_score=[prob[i] for prob in probabilities])
 
-            confusion_matrices = self._get_CMs(i, probabilities, references, thresholds)
-
-            low_ci_tpr, high_ci_tpr = [0] * len(thresholds[i]), [0] * len(thresholds[i])
-            λ = 1.0
-            for k in range(len(thresholds[i])):
-                variates = numpy.random.beta(confusion_matrices[k]["TP"] + λ, confusion_matrices[k]["FN"] + λ, 1000000)
-                low_ci_tpr[k], high_ci_tpr[k] = self._evaluate_statistics(variates, 0.95)
-
-            roc_auc_ci_low[i] = auc(fpr[i], low_ci_tpr)
-            roc_auc_ci_high[i] = auc(fpr[i], high_ci_tpr)
-
+            if CI:
+                confusion_matrices = self._get_CMs(i, probabilities, references, thresholds)
+                
+                low_ci_tpr, high_ci_tpr = [0] * len(thresholds[i]), [0] * len(thresholds[i])
+                λ = 1.0
+                for k in range(len(thresholds[i])):
+                    variates = numpy.random.beta(confusion_matrices[k]["TP"] + λ, confusion_matrices[k]["FN"] + λ, 1000000)
+                    low_ci_tpr[k], high_ci_tpr[k] = self._evaluate_statistics(variates, 0.95)
+    
+                roc_auc_ci_low[i] = auc(fpr[i], low_ci_tpr)
+                roc_auc_ci_high[i] = auc(fpr[i], high_ci_tpr)
 
 
             roc_auc[i] = auc(fpr[i], tpr[i])
@@ -123,12 +123,13 @@ class AVG_MULTICLASS_AUROC(evaluate.Metric):
 
         # Compute average AUC
         average_auc = numpy.mean(list(roc_auc.values()))
-        average_auc_ci_low = numpy.mean(list(roc_auc_ci_low.values()))
-        average_auc_ci_high = numpy.mean(list(roc_auc_ci_high.values()))
+        if CI:
+            average_auc_ci_low = numpy.mean(list(roc_auc_ci_low.values())) 
+            average_auc_ci_high = numpy.mean(list(roc_auc_ci_high.values()))
 
         return {
             "mc_auroc_score": average_auc,
-            "mc_auroc_ci": (average_auc_ci_low, average_auc_ci_high)
+            "mc_auroc_ci": (average_auc_ci_low, average_auc_ci_high) if CI else None
         }
 
     def _get_CMs(self, i, probabilities, references, thresholds):