Update main.py

AUC Updated

benchmark/runner/main.py

@@ -113,6 +113,8 @@ def normalize_probabilities(scores):
 
 # Function to calculate accuracy
 def calculate_accuracy(y_pred, labels):
+    y_pred = (y_pred - y_pred.min(axis=0)) / (y_pred.max(axis=0) - y_pred.min(axis=0) + 1e-10)
+    y_pred = np.array([[value, 1- value] for value in y_pred])
     y_pred_label = np.argmax(y_pred, axis=1)
     correct = np.sum(labels == y_pred_label)
     accuracy = 100 * correct / len(y_pred)
@@ -121,11 +123,14 @@ def calculate_accuracy(y_pred, labels):
 
 # Function to calculate AUC
 def calculate_auc(y_pred, labels, n_classes):
-    thresholds = np.arange(0.0, 1.01, .01)
+    # Normalize y_pred for each class
+    y_pred = (y_pred - y_pred.min(axis=0)) / (y_pred.max(axis=0) - y_pred.min(axis=0) + 1e-10)
+    y_pred = np.array([[value, 1- value] for value in y_pred])
+    thresholds = np.arange(0.0, 1.01, 0.01)
     fpr = np.zeros([n_classes, len(thresholds)])
     tpr = np.zeros([n_classes, len(thresholds)])
     roc_auc = np.zeros(n_classes)
-
+    
     for class_item in range(n_classes):
         all_positives = sum(labels == class_item)
         all_negatives = len(labels) - all_positives
@@ -146,13 +151,14 @@ def calculate_auc(y_pred, labels, n_classes):
         fpr[class_item, 0] = 1
         tpr[class_item, 0] = 1
         for threshold_item in range(len(thresholds) - 1):
-            roc_auc[class_item] += .5 * (tpr[class_item, threshold_item] + tpr[class_item, threshold_item + 1]) * (
+            roc_auc[class_item] += 0.5 * (tpr[class_item, threshold_item] + tpr[class_item, threshold_item + 1]) * (
                         fpr[class_item, threshold_item] - fpr[class_item, threshold_item + 1])
 
     roc_auc_avg = np.mean(roc_auc)
     print(f"Simplified average ROC AUC = {roc_auc_avg:.3f}")
     return roc_auc
 
+
 # Summarize results
 def summarize_result(result):
     num_correct_files = 0
@@ -227,4 +233,4 @@ def summarize_result(result):
         "mode": args.mode
     }
     result = run_test(**config)
-    summarize_result(result)
+    summarize_result(result)