updated AMI documentation for issue scikit-learn#2686

.. _clustering:

Author: jess010

sklearn/metrics/cluster/supervised.py

@@ -596,8 +596,11 @@ def adjusted_mutual_info_score(labels_true, labels_pred):
     Returns
     -------
     ami: float(upperlimited by 1.0)
-       The AMI takes a value of 1 when the two partitions
-       are identical and a value of 0 is expected for random partitions.
+       The AMI returns a value of 1 when the two partitions
+       are identical (ie perfectly matched), and a value of 0 is expected 
+       for random partitions. A negative value greater than may also 
+       be returned if the actual mutual information is less than the 
+       expectation value of the mutual information (independent labelling).
 
     See also
     --------