add a metrics for calculating pearson correlation coefficient

tensor2tensor/utils/metrics.py

@@ -43,6 +43,7 @@ class Metrics(object):
   APPROX_BLEU = "approx_bleu_score"
   RMSE = "rmse"
   LOG_POISSON = "log_poisson"
+  PEARSON = "pearson"
   R2 = "r_squared"
   ROUGE_2_F = "rouge_2_fscore"
   ROUGE_L_F = "rouge_L_fscore"
@@ -741,6 +742,20 @@ def from_characters(raw, lookup_):
     return distance / reference_length, reference_length
 
 
+def pearson_correlation_coefficient(predictions, labels, weights_fn=None):
+  """Calculate pearson correlation coefficient.
+
+  Args:
+    predictions: The raw predictions.
+    labels: The actual labels.
+    weights_fn: Weighting function.
+
+  Returns:
+    The pearson correlation coefficient.
+  """
+  _, pearson = tf.contrib.metrics.streaming_pearson_correlation(predictions, labels)
+  return pearson, tf.constant(1.0)
+
 # Metrics are functions that take predictions and labels and return
 # a tensor of metrics and a tensor of weights.
 # If the function has "features" as an argument, it will receive the whole
@@ -756,6 +771,7 @@ def from_characters(raw, lookup_):
     Metrics.APPROX_BLEU: bleu_hook.bleu_score,
     Metrics.RMSE: padded_rmse,
     Metrics.LOG_POISSON: padded_log_poisson,
+    Metrics.PEARSON: pearson_correlation_coefficient,
     Metrics.R2: padded_variance_explained,
     Metrics.ROUGE_2_F: rouge.rouge_2_fscore,
     Metrics.ROUGE_L_F: rouge.rouge_l_fscore,

tensor2tensor/utils/metrics_test.py

@@ -319,6 +319,21 @@ def testMultilabelMatch3(self):
       actual = session.run(a)
     self.assertAlmostEqual(actual, expected, places=6)
 
+  def testPearsonCorrelationCoefficient(self):
+    predictions = np.random.rand(12, 1)
+    targets = np.random.rand(12, 1)
+
+    expected = np.corrcoef(np.squeeze(predictions), np.squeeze(targets))[0][1]
+    with self.test_session() as session:
+      pearson, _ = metrics.pearson_correlation_coefficient(
+              tf.constant(predictions, dtype=tf.float32), 
+              tf.constant(targets, dtype=tf.float32))
+      session.run(tf.global_variables_initializer())
+      session.run(tf.local_variables_initializer())
+      actual = session.run(pearson)
+    print(actual)
+    print(expected)
+    self.assertAlmostEqual(actual, expected)
 
 if __name__ == '__main__':
   tf.test.main()