Merge pull request scikit-learn#983 from GaelVaroquaux/faster_ward

MRG: Faster ward

Author: agramfort

doc/tutorial/statistical_inference/unsupervised_learning.rst

@@ -214,7 +214,7 @@ transposed data.
    >>> agglo = cluster.WardAgglomeration(connectivity=connectivity,
    ...                                   n_clusters=32)
    >>> agglo.fit(X) # doctest: +ELLIPSIS
-   WardAgglomeration(connectivity=...
+   WardAgglomeration(compute_full_tree='auto',...
    >>> X_reduced = agglo.transform(X)
 
    >>> X_approx = agglo.inverse_transform(X_reduced)

doc/whats_new.rst

@@ -40,6 +40,11 @@ Changelog
 
    - A common testing framework for all estimators was added.
 
+   - Speedups in hierarchical clustering by `Gael Varoquaux`_. In
+     particular building the tree now supports early stopping. This is
+     useful when the number of clusters is not small compared to the
+     number of samples.
+
 API changes summary
 -------------------
 
@@ -62,6 +67,10 @@ API changes summary
    - All ``Base`` classes are now abstract meta classes so that they can not be
      instantiated.
 
+   - :func:`cluster.ward_tree` now also returns the parent array. This is
+     necessary for early-stopping in which case the tree is not
+     completely built.
+
 .. _changes_0_11:
 
 0.11