Corrected f_beta computation (#4183)

* Update f_beta.py Added METRIC_EPS in the denominator to avoid nan values in f_beta score. * Update f_beta.py Made changes flake8 compliant * Update f_beta.py Makes use of class_reduce for macro f_beta computation to avoid nans * Update f_beta.py Made flake8 compliant * Corrected F beta computation * Removed offset to make the computation precise
Lightning-AI · Oct 21, 2020 · 5d1583d · 5d1583d
1 parent a4fa7f8
commit 5d1583d
Show file tree

Hide file tree

Showing 3 changed files with 19 additions and 7 deletions.
diff --git a/pytorch_lightning/metrics/classification/f_beta.py b/pytorch_lightning/metrics/classification/f_beta.py
@@ -21,8 +21,8 @@
 import torch
 from torch import nn
 from pytorch_lightning.metrics.metric import Metric
+from pytorch_lightning.metrics.functional.reduction import class_reduce
 from pytorch_lightning.metrics.classification.precision_recall import _input_format
-from pytorch_lightning.metrics.utils import METRIC_EPS
 
 
 class Fbeta(Metric):
@@ -121,12 +121,14 @@ def compute(self):
         Computes accuracy over state.
         """
         if self.average == 'micro':
-            precision = self.true_positives.sum().float() / (self.predicted_positives.sum() + METRIC_EPS)
-            recall = self.true_positives.sum().float() / (self.actual_positives.sum() + METRIC_EPS)
+            precision = self.true_positives.sum().float() / (self.predicted_positives.sum())
+            recall = self.true_positives.sum().float() / (self.actual_positives.sum())
 
-            return (1 + self.beta ** 2) * (precision * recall) / (self.beta ** 2 * precision + recall)
         elif self.average == 'macro':
-            precision = self.true_positives.float() / (self.predicted_positives + METRIC_EPS)
-            recall = self.true_positives.float() / (self.actual_positives + METRIC_EPS)
+            precision = self.true_positives.float() / (self.predicted_positives)
+            recall = self.true_positives.float() / (self.actual_positives)
 
-            return ((1 + self.beta ** 2) * (precision * recall) / (self.beta ** 2 * precision + recall)).mean()
+        num = (1 + self.beta ** 2) * precision * recall
+        denom = self.beta ** 2 * precision + recall
+
+        return class_reduce(num=num, denom=denom, weights=None, class_reduction='macro')
diff --git a/tests/metrics/classification/inputs.py b/tests/metrics/classification/inputs.py
@@ -35,6 +35,14 @@
     target=torch.randint(high=2, size=(NUM_BATCHES, BATCH_SIZE, NUM_CLASSES))
 )
 
+# Generate edge multilabel edge case, where nothing matches (scores are undefined)
+__temp_preds = torch.randint(high=2, size=(NUM_BATCHES, BATCH_SIZE, NUM_CLASSES))
+__temp_target = abs(__temp_preds - 1)
+
+_multilabel_inputs_no_match = Input(
+    preds=__temp_preds,
+    target=__temp_target
+)
 
 _multiclass_prob_inputs = Input(
     preds=torch.rand(NUM_BATCHES, BATCH_SIZE, NUM_CLASSES),

diff --git a/tests/metrics/classification/test_f_beta.py b/tests/metrics/classification/test_f_beta.py
@@ -14,6 +14,7 @@
     _multidim_multiclass_inputs,
     _multidim_multiclass_prob_inputs,
     _multilabel_inputs,
+    _multilabel_inputs_no_match,
     _multilabel_prob_inputs,
 )
 from tests.metrics.utils import NUM_CLASSES, THRESHOLD, MetricTester
@@ -87,6 +88,7 @@ def _sk_fbeta_multidim_multiclass(preds, target, average='micro', beta=1.0):
         (_binary_inputs.preds, _binary_inputs.target, _sk_fbeta_binary, 1, False),
         (_multilabel_prob_inputs.preds, _multilabel_prob_inputs.target, _sk_fbeta_multilabel_prob, NUM_CLASSES, True),
         (_multilabel_inputs.preds, _multilabel_inputs.target, _sk_fbeta_multilabel, NUM_CLASSES, True),
+        (_multilabel_inputs_no_match.preds, _multilabel_inputs_no_match.target, _sk_fbeta_multilabel, NUM_CLASSES, True),
         (_multiclass_prob_inputs.preds, _multiclass_prob_inputs.target, _sk_fbeta_multiclass_prob, NUM_CLASSES, False),
         (_multiclass_inputs.preds, _multiclass_inputs.target, _sk_fbeta_multiclass, NUM_CLASSES, False),
         (